Can a factual approach using contextualized complex information provide clinical insights for appropriate transfusion at emergency departments?

Can a factual approach using contextualized complex information provide clinical insights for appropriate transfusion at emergency departments? | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Can a factual approach using contextualized complex information provide clinical insights for appropriate transfusion at emergency departments? Frédéric Garban, Christophe Cancé, Dimitri Sourd, Guillaume Dupouy, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7744688/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 9 You are reading this latest preprint version Abstract Background The proper management of blood and labile blood products (LBP) is an important public health issue. Beside surgery, traumatology and intensive care units, as well as hematological clinics, LBP can be delivered in emergency departments (ED). The circumstances and realization of these transfusions are however ill-known and difficult to compile. Yet such information could help improve LBP management. The aim of the TRANSFUSIA study, presented here was to develop and implement an artificial intelligence approach, interpretable by medical experts, for collecting and analyzing all information relative to transfusions in EDs in a single center. Methods In order to evaluate this activity, a Knowledge Hypergraph was designed to synthesize the meaning of facts and numerous relationships of interest inside the datawarehouse. Its purpose was to clearly define the complex graph of structured information, organized into a network used to contextualize and produce support information of interest for clinical decision-making. Clinical and transfusion information from the years 2020 and 2021 was gathered and computed into contextualized variables of interest. TRANSFUSIA thus collected 705 transfusions from 657 patients in Grenoble University Hospital (France).Data generated were statistically analyzed. Results The main aim was to analyze conditions that had led to transfusion in ED, evaluate the amount of LBP used, examine early mortality in the 4 following months and find elements to limit transfusion at ED. TRANSFUSIA confirmed that chronic anemia and coagulation anomalies were frequent indications for emergency LBP administration and that 25% of the patients died early after transfusion. These early deaths were linked to previous anticancer treatment or abnormal prothrombin time ratios and were also impacted by age and sex. Conclusion These results indicate a significant degree of over-transfusion in ED which could be alleviated by better prior monitoring and care of both low hemoglobin levels (to detect iron deficiency) and coagulation disorders. Additionally, the Knowledge Hypergraph- based method, developed for the TRANSFUSIA study, could be used to help decision-making facing possible LBP requirement for individual patients. Emergency medical services Blood component transfusion Artificial intelligence Anemia Coagulation disorder Figures Figure 1 Figure 2 Figure 3 Introduction Managing the proper adequation between blood resources and delivery is a daily public health challenge in transfusion medicine (TM). A range of good practice recommendations and guidelines (latest recommendations of the Association for the Advancement of Blood and Biotherapies in 2023 [ 1 ]), as well as other studies, have been published to improve the appropriateness of blood transfusion [ 2 – 4 ]. The Frankfurt consensus [ 5 ] for surgery and anesthesiology was the outcome of years of work in patient blood management (PBM). In internal medicine, outside PBM and intensive care, national and international recommendations are not easy to apply, owing to the diversity of clinical situations, with a trend in favor of restrictive policies [ 6 , 7 ]. There are two major fields in TM. First, blood transfusion in patients with hematological diseases, an area of major blood consumption and economic impact, where the best use of blood support requires individual approaches. The second field is that of emergency departments (ED) receiving patients in urgent need of transfusion, a situation that could have been better managed beforehand. Indeed, avoiding inappropriate transfusion could be achieved by anticipating deep anemia. The major causes of severe anemia in ED have been identified as iron deficiency, gastro intestinal bleeding, cancer history and hematologic diseases [ 8 ]. Peyrony et al. [ 8 ], in a study of ED transfusions, reported 40.2% of patients with chronic anemia, and absence of medical motive for transfusion in 39.9%. Paganini et al. [ 9 ], in a retrospective study, also observed that 40.4% of transfusions in ED were inappropriate. Although criteria used to define severe anemia and transfusion thresholds, i.e. mostly low levels of hemoglobin (Hb), do not differ widely, it is interesting to notice that chronic or multifactorial anemia was found to be present in about 20% of patients requiring blood support in ED [ 8 ]. The use of antithrombotic drugs is also well documented as a contributing factor for bleeding and hence transfusion in ED [ 8 – 10 ]. Consequently, a significant proportion of red blood cell (RBC) transfusions in ED could be considered inappropriate in the sense of “over transfusion” [ 8 – 11 ]. This notion could be even more valid if these transfusions are followed by early death. The main objective of the study reported here was to evaluate the early death, within four months after transfusion at ED, a suspected but ill-documented incidence. Better knowledge on this matter could help for a better PBM. To this avail, instead of setting-up a clinical research protocol, a broad retrospective survey of blood transfusions in ED was performed in Grenoble Alpes University Hospital, through an innovative method based on real-life information and artificial intelligence (AI). The TRANSFUSIA project automatically collected retrospective data from patient electronic health records (EHR), and used an approach combining Knowledge Hypergraph [ 12 – 14 ] networks to process raw data and create medical information of interest. Specifically, anemia and coagulation disorders were explored together with the management and outcome of patients transfused in ED. Methods Overview –ethical considerations The TRANSFUSIA project was a research program focusing on understanding the processes leading to an indication of transfusion, through the health (HDW) and clinical (CDW) data warehouses of Grenoble University Hospital, called PREDIMED (Plateforme de Rassemblement et d'Exploitation des Données bIoMEDicales). All patients included in the TRANSFUSIA study were informed. According to the French law allowing for research on health data, formal consent is not required , only opposition of patients is formal [15]. Therefore, data from patients who refused that their information was used were removed (< 3%). Conformity according to French law was reviewed by the institutional Ethic committee of the university of Grenoble Alpes (Institution review board). This project was approved by the ethics committee of Grenoble University (CERGA) on march 23, 2023 (CERGA-Avis-2023-08). This study was conducted according to the Declaration of Helsinki. Knowledge Hypergraph approach and main features Health data are highly connected as the link between data and information is at the basis of medical reasoning. The structure of information from CDW, based on Knowledge Hypergraph [13], describes the organization of a highly connected network. The latter contains classes and hyperclasses that represent sets of “objects” and potential semantic or topological relationships between these “objects” and their “attributes”. This graphic conceptual model is flexible and suited to specify and implement the contextualization of highly connected data network elements (Figure. This method allows answering potentially complex problems through successive simple graph transversal queries [13]. First level data, or elaborated information such as medical events, can then be contextualized in relation to each other, especially temporally, even if they are structurally distant in the information structure. Moreover, Knowledge Hypergraph modeling allows preserving the granularity of the original information. Finally, graphs networked according to the Knowledge Hypergraph model are suitable for an interactive exploration of information, combined to build trajectories describing the longitudinal follow-up of patients [14, 16-19]. Data extraction and elaboration De-identified data from the PREDIMED CDW included administrative and demographic information, hospitalization details, laboratory results, diagnoses, procedures, medications and free-text documents from each patient EHR. A specific data extraction from the CDW [13] was used for TRANSFUSIA, including all adult patients admitted to Grenoble Alpes University Hospital for anemia-related transfusion between 2020 and 2021. Patients were identified by crossing medical administrative data from the medical information department and medical data from the LBP registry. All patients over 18 years of age, referred to the hospital for transfusion during a short or emergency stay were considered. The first level data-set consisted of all information before and after transfusion (Supplemental material). Also recorded was all relevant information in French public epidemiology registries and administrative data. Raw data such as age, Hb level, CBC, and any numeric biological value or simple laboratory results (e.g. positive/negative), were processed to build a set of categorical variables. Categories were designed according to guidelines defining iron deficiency, i.e. ferritin level < 30 mg/L (or < 45 mg/L for patients over 60-year-old) and/or transferrin saturation < 25 % [20]. Second level: enrichment of the structured data network with unstructured data Antithrombotic treatments are often discontinued upon admission in ED, with no trace in the dataset of hospital prescription. However, doctors systematically report outpatient treatments in the initial textual medical record. An algorithm was thus built to detect any antithrombotic drug prescription in medical records of full-text ED admission reports. Finally, this new categorical variable, constituting a second level of elaborated information, was used to enrich the Knowledge Hypergraph . The final structured data network with clinically relevant information was constructed through the Knowledge Hypergraph data network as an innovative approach using graph transversal queries on data organized as a network (Figure 1). The first event of interest was the “transfusion episode in ED”. To get a complete view of transfusion events, they were considered as a treatment of severe anemia of medical origin, However, episodes of acute anemia were not necessarily resolved in the ED, and LBP may thus have been administered both in the ED and in another hospital unit. A period of no more than 3 days was considered during which one or more LBP were administered to patients in ED. Each episode was defined by the number and type of LBP dispensed, dates of first and last administration, and their relative duration. In the Knowledge Hypergraph model, the new class “transfusion episode” was created within the “treatment” hyperclass. Simple graphical queries were then used to explore the clinical context of transfusion episodes. Based on this principle, the “emergency transfusion” event was temporally updated with the following context: list of pre-transfusion laboratory tests list of pre-transfusion administered treatments list of pre-transfusion ICD10 Z51 diagnosis codes [21]. The final data model is depicted in Figure 2. More detailed information is provided in Supplemental material. Quality control Before analyses were initiated, each set of extracted variables was screened for coherence with a pluridisciplinary check involving medical computer scientists, biostatisticians, and medical experts. Improbable values for numeric data were checked and amended or removed. Mistakes in encoding the ED stay were detected and corrected. Concerning “transfusion at ED”, there was no encoding nor typing mistake, since all LBP and whole transfusion process are electronically managed. This ensured an automatic link between the blood bank and hospital information system. All medical reports were reviewed by two hematologists to define the causes of anemia that had led to transfusion at ED. These findings were not analyzed as variables, but provided a quality check that turned out to be satisfactory. Finally, the controlled dataset was frozen before initiating statistical analyses according to local procedures following STROBE (https://www.equator-network.org/reporting-guidelines/strobe/) clinical research procedures. Statistics Categorical variables were described as frequencies and percentages, with exact confidence intervals. Continuous variables were summarized by medians and interquartile ranges (IQR). Factors associated with early death were analyzed using Fisher exact test for categorical data. For continuous variables, comparisons were made using Student t-test for normally distributed data, or Wilcoxon-Mann-Whitney test. The normality of continuous variables was assessed through graphical inspection. All statistical analyses were conducted using R software version 3.5.2. A significance threshold of 0.05 was applied for P- values. Results Transfusion characteristics With a starting point defined as an emergency blood transfusion episode for serious anemia in ED, evaluated within the consecutive 3 days after transfusion, 705 transfusions episodes were recorded, representing 1818 RBC units, 81 platelet units and 61 plasma units. The vast majority of transfusions required 2 LBP (median 2, IQR 2-3). Transfusions requiring 4 LBP or more were considered as related to active bleeding and secondarily relevant to intensive care unit (massive bleeding or shock) or gastro enterology department in case of digestive bleeding (Table 1). Table 1 Transfusion characteristics according to episodes and outcome. Univariate and multivariate analyses Univariate analysis (n=705) Multivariate analysis* (n=671) Transfusions N=705 All n=705 Death < 4 months n=173 Alive at 4 months n=532 Odds ratio (95%CI) p-value Odds ratio (95%CI) P -value Age years median (IQR) 80 (66-87) 83 (72-89) 78 (63-86) 1.03 (1.02-1.04) <0.001 1.03 (1.02-1.05) <0.001 Sex (women) , n (%) 325 (46.1) 68 (39.3) 257 (48.3) 0.69 (0.49-0.98) 0.04 0.59 (0.39-0.88) 0.009 Red blood cell consumption, Units median (IQR) 2 (2-3) 2 (2-3) 2 (2-3) 0.87 (0.31-1.00) 0.05 0.77 (0.64-0.91) 0.002 Platelet concentrate consumption, Units median (IQR) 0 (0-0) 0 (0-0) 0 (0-0) 1.21 (0.91-1.58) 0.17 Plasma concentrate consumption, Units median (IQR) 0 (0-0) 0 (0-0) 0 (0-0) 1.02 (0.80-1.23) 0.83 Hemoglobin, g/L N=672 median (IQR) 74.3 (16.8) 75.3 (16.3) 74.0 (16.9) 1.00 (0.99-1.01) 0.47 Leukocytes x10 9 /L N=672 median (IQR) 7.7 (5.4-11.3) 9.2 (6.2-16.1) 7.4 (5.3-10.2) 1.05 (1.03-1.08) <0.001 1.04 (1.02-1.07) 0.002 Neutrophils, x10 9 /L N=520 median (IQR) 5.3 (3.7-6.9) 6.4 (3.9-11.3) 5.2 (3.7-7.5) 1.08 (1.04-1.12) <0.001 Platelet count, x10 9 /L N=669 median (IQR) 227 (146-334) 191 (112-299) 239 (162-342) 0.99 (0.99-0.99) 0.002 Platelets < 50x10 9 /L n (%) 39 (5.5) 17 (9.8) 22 (4.1) 2.53 (1.32-4.83) 0.006 Prothrombin ratio <70%, n (%) 256 (36.3) 94 (54.3) 162 (30.5) 2.72 (1.91-3.86) <0.001 2.43 (1.64-3.61) =1.2, n (%) 326 (46.2) 109 (63.0) 217 (40.8) 2.47 (1.74-3.52) 1.2, n (%) 238 (33.8) 83 (48.0) 155 (29.1) 2.24 (1.58-3.19) <0.001 Anti-thrombotic, n (%) 253 (35.9) 67 (38.7) 186 (35.0) 1.18 (0.83-1.67) 0.41 Iron injection (<1y) n (%) 221 (31.3) 53 (30.6) 168 (31.6) 0.96 (0.66-1.39) 0.85 Iron deficiency, n (%) 263 (37.3) 50 (28.9) 213 (40.0) 0.61 (0.42-0.88) 0.01 0.51 (0.33-0.77) 0.002 Anti-cancer treatment before transfusion in ED, n (%) 132 (18.7) 54 (31.2) 78 (14.7) 2.64 (1.76-3.94) <0.001 2.95 (1.83-4.78) <0.001 GI hemorrhage, N=657 n (%) 89 (13.5) 16 (9.7) 73(14.8) 0.64 (0.36-1.13) 0.12 RBC distribution index median (IQR) 17.2 (15.7-19.8) 17.2 (15.3-19.5) 17.2 (15.8-20.1) 0.96 (0.91-1.01) 0.10 CI: confidence interval; IQR: interquartile range; INR: International Normalized Ratio; APTT: activated partial thromboplastin time. ED: emergency department. GI: gastro-intestinal. *A logistic multiple, backward regression tested independent association to early death. The model included all variables with a P-value < 0.1 in univariable analysis Causes of anemia reviewed by experts Medical record review allowed for a precise definition of the causes of anemia that led to transfusion. The main cause of transfusion in ED was confirmed to be low Hb, followed by digestive tract bleeding or liver disease (cirrhosis), hematologic disease and solid tumor. In the population presenting with early death post-transfusion, there was an over-representation of hematological malignancies and less iron deficiencies. Interestingly, the algorithmic approach identified a higher frequency of iron deficiency than experts (41 % vs 27%). However, about 60 % of anemia cases were associated to two or more causes, including iron deficiency or chronic bleeding. Time dimension and patient outcome The 705 transfusion episodes concerned 657 patients admitted at ED with anemia (emergencies relevant to trauma, cardiology, surgery or intensive care were specifically managed outside the ED and not included in the study). Taking into account the time dimension, 364 (55.4%) of these 657 patients died, 173 (26.3%) of them within 4 months after emergency transfusion. COVID positive tests at any date within the period of analysis were found in only 20 patients, representing 3 % of early deaths, yet a sensitivity analysis ruled-out any COVID effect. Table 1 compares the characteristics of the 705 episodes of patients with early post-transfusion death (n=173) and alive patients at 4 months post-transfusion (n=532), providing univariate and multivariate analyses. Univariate analysis points out, in the early death population, the significance of older age and male sex. Biological parameters segregating this population are significantly higher leukocyte and neutrophil counts, lower platelet counts with more patients having less than 50x10 9 /L platelets, yet no difference in Hb levels. There were also significantly more altered coagulation tests in the early death population. Multivariate analysis confirmed that early death was significantly associated with age, male sex, number of LBP received, leukocytosis, and, more surprisingly, with a prothrombin ratio <70% and previous systemic anti-cancer therapy. In table 2, the main causes of anemia prompting transfusion are shown, 17% being related to a malignant condition. Table 2 Main causes of anemia leading to transfusion, by decreasing order Causes N (%) Iron deficiency 193 (27%) Digestive tract bleeding and liver disease 154 (22%) Chronic unknown 107 (15%) Other 83 (12%) Hematological malignancy 63 (9%) Solid tumor 56 (8%) Post bleeding anemia (iatrogenic or other) 28 (4%) Non-malignant hematological condition 21 (3%) Figure 3 provides the amount of LBP used by episode, showing the predominant use of 2 RBC bags. Discussion In line with previous publications of prospective [ 8 ] and retrospective [ 9 ] studies, the Knowledge Hypergraph approach used here allowed sorting out the main etiologies of anemia leading to emergency transfusion at ED, respectively iron deficiency and digestive bleeding. This study also disclosed that a higher number of LBP was transfused in ED (median 2 instead of the expected 1 as shown in Fig. 3 ) than in reported studies [ 8 , 9 ]. Of interest, TRANSFUSIA also highlighted an unexpected 25% rate of early deaths in this ED-transfused population, raising questioning about the rationale of these transfusions. Hb levels were clearly not a discriminating factor since all transfused patients had Hb levels justifying transfusion. Yet, other factors should clearly have been considered before initiating transfusion. Older patients with cancer, iron deficiency and coagulation anomalies could certainly have been taken care of by other means (IV iron transfusion, leveraged anti-coagulation) and their age and life expectancy considered before urging transfusion, with probably excessive product amounts. Since transfusions are clearly and rightly decided upon Hb levels, prior interventions to increase Hb in such nearly-terminal patients should be implemented before sending them to ED. The Knowledge Hypergraph -driven data elaboration used in this study provided robust results, in a different and lighter context than that of a cumbersome and costly clinical trial. It allowed to disclose a number of ill-documented and somehow unexpected results about PBM strategies in ED. The innovative Knowledge Hypergraph strategy used here could lead to new developments in the evaluation of clinical practice. Complementary to clinical studies using manually collected data, this methodology provides highly accurate audited datasets and allowed reaching an exhaustive view of transfusion at ED. Initial results of the TRANSFUSIA project, currently being mainly descriptive and partially supported by the literature, are largely consistent with recent similar studies published on the subject [ 11 , 22 , 23 ]. These findings confirm the robustness of the data acquisition method used here. The primary innovation of TRANSFUSIA was to use PREDIMED HDW for data collection. While randomized controlled trials remain the undisputed Gold Standard of clinical research, initial results from TRANSFUSIA offer a glimpse of research via CDWs. Indeed, they constitute a reliable alternative to traditional methods for non-interventional retrospective and prospective cohort projects, provided that the retrieved data is linked to the in-hospital records of recruited patients. However, further studies are needed to validate these initial findings. The Knowledge Hypergraph approach provided a multidimensional view in terms of context of each transfusion episode that preserved logical links with patient and drugs, yet allowing for the addition of algorithmic research from non-structured information (text in medical records). Nonetheless, it took analysis a step further by adding relative time information between all elements. This is a foundation encouraging the application of further neuronal approaches in order to make predictive requirements or appropriate use of blood products. Basic data such as Hb, date of transfusion, blood products, drug administration and medical reports were here exploited layer by layer, without any degradation. At the more complex layers, there was only an addition of links between primary data, and of time-related variables. This allowed considering variability with time, required for full data exploitation. Altogether, the basic data was preserved and exploited taking into account dynamics and contextualization, thus bringing more confidence in the events studied [ 19 ]. Moreover, the advantage of HDWs lies in their ability to frame validation processes for data derived from multiple sources or over long time periods, a process that would require significant human resources using traditional data collection methods. Limitations A usual limitation of studies using CDWs lies in the current question surrounding real world health data, namely quality [ 24 ]. As data from CDWs is initially collected for patient care, this has already been identified as one of the areas of improvement in CDWs, in close relationship with the notion of “fitness for use” [ 25 ]. These constraints are theoretically less of an issue in prospective studies but are inherent to retrospective cohort studies [ 26 ]. However, this limitation had little impact in the model applied here, considering the innovative Knowledge Hypergraph approach that allows building robust contextualized information from raw data. Conclusion By combining a Knowledge Hypergraph -driven data elaboration method and statistical analysis, the innovative approach used here provided new results for the understanding of transfusion at ED and some elements for patient management before this decision. It also provided a way of thinking about ethical views of transfusion, considering the high proportion of patients dying within 4 months after a transfusion at ED. Declarations Ethics approval and consent to participate This project was approved by the ethics committee (IRB) of Grenoble University (CERG) and was conducted in compliance with the Declaration of Helsinki. Consent for publication Patients were informed of the study and none opposed to the research (as per French law) Availability of data and materials Data from this study can be made available upon reasonable request from the corresponding author after authorization from Grenoble University Hospital. Competing interests The authors declare no conflict of interest related to this work. Funding This work was supported in part by institutional funding from Pfizer, Novartis, the Société Francophone de Transfusion Sanguine, ART. Authors' contributions FG designed the study, EV and FG performed extractions, CC, PAB, GD and FG developed the software, DS, AV and JLB performed statistical analyses, FG, DV, SDT, AMG and JLB wrote and edited the manuscript. Acknowledgements Medical writing for this manuscript was assisted by MPIYP (MC Béné), Paris, France. References Carson JL, Stanworth SJ, Guyatt G, Valentine S, Dennis J, Bakhtary S, et al. Red blood cell transfusion: 2023 AABB International Guidelines. JAMA. 2023;3301892-1902. Staples S, Evans H, Caulfield J, Bend M, Foy R, Murphy MF, et al. Opportunities to improve feedback to reduce blood component wastage: Results of a national scheme evaluation. Transfusion. 2024;641223-32. Lee TC, Almeida N, Pelletier P, McDonald EG. Comparison of blood transfusion rates before and after implementation of a quality improvement initiative for transfusion safety and appropriateness. JAMA Netw Open. 2023;6:e2252253. Stanworth SJ, Walwyn R, Grant-Casey J, Hartley S, Moreau L, Lorencatto F, et al. Effectiveness of enhanced performance feedback on appropriate use of blood transfusions: a comparison of 2 cluster randomized trials. JAMA Netw Open; 2022;5:e220364. Mueller MM, Van Remoortel H, Meybohm P, Aranko K, Aubron C, Burger R, et al. Patient blood management: recommendations from the 102018 Frankfurt consensus conference. JAMA. 2019;321:983-97. Moras E, Abbott JD, Vallabhajosyula S. AABB recommends restrictive RBC transfusions for hospitalized adults and children. Ann Intern Med. 2024;177:JC14. Radford M, Estcourt LJ, Sirotich E, Pitre T, Britto J, Watson M, et al. Restrictive versus liberal red blood cell transfusion strategies for people with haematological malignancies treated with intensive chemotherapy or radiotherapy, or both, with or without haematopoietic stem cell support. Cochrane Database Syst Rev. 2024; 5:CD011305. Peyrony O, Gamelon D, Brune R, Chauvin A, Ghazali DA, Yordanov Y, et al. Red blood cell transfusion in the emergency department: an observational cross-sectional multicenter study. J Clin Med. 2021;10:2475. Paganini M, Rigon F, Rebustello F, Cianci V, Bertozzi I, Randi ML. Appropriateness of packed red blood cells transfusions in chronic anemic patients in the emergency department: the TRANSFUS-ED retrospective analysis. Intern Emerg Med. 2023;18:1815-21. Bouget J, Balusson F, Viglino D, Roy PM, Lacut K, Pavageau L, et al. Major bleeding risk and mortality associated with antiplatelet drugs in real-world clinical practice. A prospective cohort study. PLoS One. 2020;15:e0237022. Green L, Tan J, Morris JK, Alikhan R, Curry N, Everington T, et al. A three-year prospective study of the presentation and clinical outcomes of major bleeding episodes associated with oral anticoagulant use in the UK (ORANGE study). Haematologica. 2018;103:738-45. Raman MRG, Somu N, Kirthivasan K, Sriram VSS. A Hypergraph and arithmetic residue-based probabilistic neural network for classification in intrusion detection systems. Neural Netw. 2017;92:89-97. Cancé C, Lenne C, Artemova S, Mossuz P, Moreau-Gaudry A. Hypergraph based data model for complex health data exploration and its implementation in PREDIMED clinical data warehouse. Stud Health Technol Inform. 2022;290:335-339. Dai, Q., Gao, Y. Hypergraph computation for medical and biological applications. In: Hypergraph Computation. Artificial Intelligence: Foundations, Theory, and Algorithms. Springer, Singapore, 2023. Boyer L, Fond G, Gauci MO, Boussat B. Regulation of medical research in France: Striking the balance between requirements and complexity. Rev Epidemiol Sante Publique. 2023;71:102126 Fatemi B, Taslakian P, Vasquez D, Poole D. Knowledge hypergraphs: prediction beyond binary relations. 2020. Twenty-Ninth International Joint Conference on Artificial Intelligence and Seventeenth Pacific Rim International Conference on Artificial Intelligence {IJCAI-PRICAI-20. DOI 0.24963/ijcai.2020/299. Last accessed September 29, 2025. Chen Z, Wang X, Wang C, Li Z. A position‑aware knowledge hypergraph model for link prediction. Data Sci Engin. 2023;8:135-45. Lamine SBAB, Radaoui M, Zghal HB. Knowledge hypergraph-based multidimensional analysis for natural language queries: application to medical data. computational science – ICCS 2023: 23rd International Conference, Prague, Czech Republic, July 3–5, 2023, Proceedings, Part III, Aydin, Berkay & Angryk, Rafal. (2018). Modeling Spatiotemporal Relationships Among Trajectories. 10.1007/978-3-319-99873-2_3. Iolascon A, Andolfo I, Russo R, Sanchez M, Busti F, Swinkels D,et al. Recommendations for diagnosis, treatment, and prevention of iron deficiency and iron deficiency anemia. Hemasphere. 2024;8:e108. Brämer GR. International statistical classification of diseases and related health problems. Tenth revision. World Health Stat Q. 1988;4132-36. Maynard S, Farrington J, Alimam S, et al. Machine learning in transfusion medicine: A scoping review. Transfusion 2024; 64:162-184 Evans HG, Murphy MF, Foy R, Evans H, Li K, Wong WK, et al. Harnessing the potential of data-driven strategies to optimise transfusion practice. Br J Haematol.2024;204:74-85. Bocquet F, Campone M, Cuggia M. The challenges of implementing comprehensive clinical data warehouses in hospitals. Int J Environ Res Public Health. 2022;19:7379. Wang RY, Strong DM. Beyond accuracy: what data quality means to data consumers. J ManInf Syst. 1996;14:5-33. Matsumoto N, Moran J, Choi H, et al. KRAGEN: a knowledge graph-enhanced RAG framework for biomedical problem solving using large language models. Bioinformatics. 2024;40:btae353. Additional Declarations No competing interests reported. Supplementary Files SupplementalTRANSFUSIABMCMIDM.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 26 Nov, 2025 Reviews received at journal 25 Nov, 2025 Reviewers agreed at journal 20 Nov, 2025 Reviewers agreed at journal 19 Nov, 2025 Reviewers invited by journal 19 Nov, 2025 Editor assigned by journal 18 Nov, 2025 Editor invited by journal 30 Oct, 2025 Submission checks completed at journal 29 Oct, 2025 First submitted to journal 29 Oct, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7744688","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":551308826,"identity":"bd330afa-6aae-410a-b8e3-db28cf05b92b","order_by":0,"name":"Frédéric Garban","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABBklEQVRIiWNgGAWjYBAC9gYGBmYgLccgAROSYAaK2eDWwnMAosUYSQsjUEsaYS2JDcRr4V/87HNBzZ30/tk9Boxf99jk8Us3Nj5gSLiHW4vEM+PZM449y51x54wBs8yztGLJOQebDRgSinFqsZc4YMzMw3Y4d4NEjvlviQOHEzfcSGyTYPyRgMeW45+Zef4dTjeQyDFgljjwH6Sl/QdDAh4t/D3GzLxthxNAWhg/HDgAtoUBrxYJnmLmmX2HDWfcSCtgZjiQnDgT6BeJBLy2HN/MXPDtsDz/jOQNjD8O2CX2Szcf/PABjxYGCSQ5Zh4YC48GBgb+Awg24w98KkfBKBgFo2DEAgANaFewncKoFAAAAABJRU5ErkJggg==","orcid":"","institution":"Grenoble-Alpes University, UMR 5525","correspondingAuthor":true,"prefix":"","firstName":"Frédéric","middleName":"","lastName":"Garban","suffix":""},{"id":551308827,"identity":"5af1039f-17b4-4bc0-ba83-00b19286878c","order_by":1,"name":"Christophe Cancé","email":"","orcid":"","institution":"Grenoble-Alpes University, UMR 5525","correspondingAuthor":false,"prefix":"","firstName":"Christophe","middleName":"","lastName":"Cancé","suffix":""},{"id":551308830,"identity":"ddb13938-c01e-4092-a379-834fc8d36859","order_by":2,"name":"Dimitri Sourd","email":"","orcid":"","institution":"Grenoble-Alpes University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Dimitri","middleName":"","lastName":"Sourd","suffix":""},{"id":551308831,"identity":"d87b52e9-8e67-4ec8-9546-771c92b5e389","order_by":3,"name":"Guillaume Dupouy","email":"","orcid":"","institution":"Grenoble-Alpes University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Guillaume","middleName":"","lastName":"Dupouy","suffix":""},{"id":551308832,"identity":"e677874e-2caf-4e6b-8266-6b02ff020725","order_by":4,"name":"Eve Laborde","email":"","orcid":"","institution":"Grenoble Alpes University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Eve","middleName":"","lastName":"Laborde","suffix":""},{"id":551308833,"identity":"6c0a8ddd-ca8a-4151-ac3e-7d0c76c8214c","order_by":5,"name":"Damien Viglino","email":"","orcid":"","institution":"Grenoble-Alpes University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Damien","middleName":"","lastName":"Viglino","suffix":""},{"id":551308834,"identity":"d477e67d-9fd5-4662-a3dd-53223d8cfec1","order_by":6,"name":"Paul Antoine Beaudoin","email":"","orcid":"","institution":"Grenoble-Alpes University, UMR 5525","correspondingAuthor":false,"prefix":"","firstName":"Paul","middleName":"Antoine","lastName":"Beaudoin","suffix":""},{"id":551308835,"identity":"9e3964ce-79b6-445e-9be0-4de1469eca2b","order_by":7,"name":"Antoine Vilotitch","email":"","orcid":"","institution":"Grenoble-Alpes University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Antoine","middleName":"","lastName":"Vilotitch","suffix":""},{"id":551308836,"identity":"d1a2aadd-c237-462d-b36f-6a9321665fab","order_by":8,"name":"Sandra David Tchouda","email":"","orcid":"","institution":"Grenoble-Alpes University Hospital","correspondingAuthor":false,"prefix":"","firstName":"Sandra","middleName":"David","lastName":"Tchouda","suffix":""},{"id":551308837,"identity":"a144db2f-f161-46a2-a3d0-dfb4fd1f7f3e","order_by":9,"name":"Alexandre Moreau Gaudry","email":"","orcid":"","institution":"Grenoble-Alpes University, UMR 5525","correspondingAuthor":false,"prefix":"","firstName":"Alexandre","middleName":"Moreau","lastName":"Gaudry","suffix":""},{"id":551308838,"identity":"1e96cf3f-7eaf-4d8c-993f-c6de4f6aedad","order_by":10,"name":"Jean Luc Bosson","email":"","orcid":"","institution":"Grenoble-Alpes University, UMR 5525","correspondingAuthor":false,"prefix":"","firstName":"Jean","middleName":"Luc","lastName":"Bosson","suffix":""}],"badges":[],"createdAt":"2025-09-29 18:23:25","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7744688/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7744688/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":97140172,"identity":"1a666503-2883-46ce-a36f-39c30c84d51c","added_by":"auto","created_at":"2025-12-01 10:03:59","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1059116,"visible":true,"origin":"","legend":"","description":"","filename":"TRANSFUSIABMCMIDM131025cd.docx","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/dfbcbcb66848bdb9642a62eb.docx"},{"id":97140413,"identity":"96c8a696-46eb-476f-a8cd-d30d81081383","added_by":"auto","created_at":"2025-12-01 10:04:57","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11965,"visible":true,"origin":"","legend":"","description":"","filename":"237d23b0d73b4f9786c6a40817c14767.json","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/44ff3a6a56ac087700576ad9.json"},{"id":97095873,"identity":"c34cf2f5-63da-4a6b-af2f-dcb71757cc9a","added_by":"auto","created_at":"2025-11-30 23:26:44","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":656030,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalTRANSFUSIABMCMIDM.docx","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/3ad75a911e6af835afcbb684.docx"},{"id":97140297,"identity":"eaf397b8-16d6-43f5-8ea0-3481c8bbbd67","added_by":"auto","created_at":"2025-12-01 10:04:31","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":95321,"visible":true,"origin":"","legend":"","description":"","filename":"237d23b0d73b4f9786c6a40817c147671enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/83ddb9ae99d8744af2262b06.xml"},{"id":97140443,"identity":"61994f36-54df-4326-bf9f-017907228728","added_by":"auto","created_at":"2025-12-01 10:05:02","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":368998,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/05450dd0eeed2be793db8b86.jpeg"},{"id":97141759,"identity":"5dbdb054-1850-4545-934b-175a1bbe87f2","added_by":"auto","created_at":"2025-12-01 10:06:59","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":504888,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/9f96ed5bbc89f9d2a00be8d5.jpeg"},{"id":97095877,"identity":"91027324-d8f3-40cc-908e-5a8b9595b387","added_by":"auto","created_at":"2025-11-30 23:26:44","extension":"jpeg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":120476,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/ab373963ad12774f3d5e11c8.jpeg"},{"id":97095875,"identity":"bc84b7bd-df30-4be2-b63b-27f002c99c34","added_by":"auto","created_at":"2025-11-30 23:26:44","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":54146,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/08df0ecbd53e339578962db3.png"},{"id":97095879,"identity":"5acf86a7-3869-4c27-bfe5-e64051ef80ff","added_by":"auto","created_at":"2025-11-30 23:26:44","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":75131,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/98f810a2e39a3d9a9371861f.png"},{"id":97095874,"identity":"99743590-c6c8-40e6-88c1-84f3cfbae007","added_by":"auto","created_at":"2025-11-30 23:26:44","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9225,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/c5ad62c821e28c7596e49787.png"},{"id":97095880,"identity":"0bb94998-869a-4a27-804c-6d0135e03c25","added_by":"auto","created_at":"2025-11-30 23:26:44","extension":"xml","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":90344,"visible":true,"origin":"","legend":"","description":"","filename":"237d23b0d73b4f9786c6a40817c147671structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/dfac878997da16d87a776aef.xml"},{"id":97095881,"identity":"da8c079d-8513-4761-86ff-97688a2bd901","added_by":"auto","created_at":"2025-11-30 23:26:44","extension":"html","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":103353,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/fee2c5d2fd17a6b12af127d1.html"},{"id":97139890,"identity":"e07705af-0304-4580-acf1-800600ad669a","added_by":"auto","created_at":"2025-12-01 10:02:59","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":371339,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDataset construction\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/678bb12e8ccdba43af4d2e4c.png"},{"id":97095867,"identity":"0cc46ac1-86cb-4520-be3c-183f7f0cf419","added_by":"auto","created_at":"2025-11-30 23:26:44","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":358344,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMultidimensional graphic representation of the specific warehouse data used in the TRANSFUSIA Study\u003c/strong\u003e. Curved arrows show the links built between each field of data. Bold rectangles point out the key-fields of data used for studying transfusion at emergency department , each click on one box or link is implemented in the address bar shown above the figure.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/be41b806f8fdd77b18bc0851.png"},{"id":97095870,"identity":"206dc46f-590b-4ece-948b-384c5edf7229","added_by":"auto","created_at":"2025-11-30 23:26:44","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":130639,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eNumber of blood products per transfusion episode\u003c/strong\u003e (RBC: red blood cells; Plt: platelets).\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/ef6e31dcd6192222afb49786.png"},{"id":97144998,"identity":"3fef2a81-6f07-4911-a05b-3c1ed69254a8","added_by":"auto","created_at":"2025-12-01 10:12:44","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1695047,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/cddb3aa1-304a-4208-afd9-ceb79fa228a8.pdf"},{"id":97095872,"identity":"72f1bf98-86ea-4872-a1e0-4e6c686f58d5","added_by":"auto","created_at":"2025-11-30 23:26:44","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":656030,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalTRANSFUSIABMCMIDM.docx","url":"https://assets-eu.researchsquare.com/files/rs-7744688/v1/305f00ec55950977109a7066.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Can a factual approach using contextualized complex information provide clinical insights for appropriate transfusion at emergency departments?","fulltext":[{"header":"Introduction","content":"\u003cp\u003eManaging the proper adequation between blood resources and delivery is a daily public health challenge in transfusion medicine (TM). A range of good practice recommendations and guidelines (latest recommendations of the Association for the Advancement of Blood and Biotherapies in 2023 [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]), as well as other studies, have been published to improve the appropriateness of blood transfusion [\u003cspan additionalcitationids=\"CR3\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. The Frankfurt consensus [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e] for surgery and anesthesiology was the outcome of years of work in patient blood management (PBM). In internal medicine, outside PBM and intensive care, national and international recommendations are not easy to apply, owing to the diversity of clinical situations, with a trend in favor of restrictive policies [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. There are two major fields in TM. First, blood transfusion in patients with hematological diseases, an area of major blood consumption and economic impact, where the best use of blood support requires individual approaches. The second field is that of emergency departments (ED) receiving patients in urgent need of transfusion, a situation that could have been better managed beforehand. Indeed, avoiding inappropriate transfusion could be achieved by anticipating deep anemia.\u003c/p\u003e\u003cp\u003eThe major causes of severe anemia in ED have been identified as iron deficiency, gastro intestinal bleeding, cancer history and hematologic diseases [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Peyrony et al. [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e], in a study of ED transfusions, reported 40.2% of patients with chronic anemia, and absence of medical motive for transfusion in 39.9%. Paganini et al. [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e], in a retrospective study, also observed that 40.4% of transfusions in ED were inappropriate. Although criteria used to define severe anemia and transfusion thresholds, i.e. mostly low levels of hemoglobin (Hb), do not differ widely, it is interesting to notice that chronic or multifactorial anemia was found to be present in about 20% of patients requiring blood support in ED [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eThe use of antithrombotic drugs is also well documented as a contributing factor for bleeding and hence transfusion in ED [\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Consequently, a significant proportion of red blood cell (RBC) transfusions in ED could be considered inappropriate in the sense of \u0026ldquo;over transfusion\u0026rdquo; [\u003cspan additionalcitationids=\"CR9 CR10\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. This notion could be even more valid if these transfusions are followed by early death.\u003c/p\u003e\u003cp\u003eThe main objective of the study reported here was to evaluate the early death, within four months after transfusion at ED, a suspected but ill-documented incidence. Better knowledge on this matter could help for a better PBM. To this avail, instead of setting-up a clinical research protocol, a broad retrospective survey of blood transfusions in ED was performed in Grenoble Alpes University Hospital, through an innovative method based on real-life information and artificial intelligence (AI). The TRANSFUSIA project automatically collected retrospective data from patient electronic health records (EHR), and used an approach combining \u003cem\u003eKnowledge Hypergraph\u003c/em\u003e [\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e] networks to process raw data and create medical information of interest. Specifically, anemia and coagulation disorders were explored together with the management and outcome of patients transfused in ED.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cem\u003eOverview \u0026ndash;ethical considerations\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe TRANSFUSIA project was a research program focusing on understanding the processes leading to an indication of transfusion, through the health (HDW) and clinical (CDW) data warehouses of Grenoble University Hospital, called PREDIMED (Plateforme de Rassemblement et d\u0026apos;Exploitation des Donn\u0026eacute;es bIoMEDicales). All patients included in the TRANSFUSIA study were informed. According to the French law allowing for research on health data, formal consent is not required , only opposition of patients is formal [15]. Therefore, data from patients who refused that their information was used were removed (\u0026lt; 3%). Conformity according to French law was reviewed by the institutional Ethic committee of the university of Grenoble Alpes (Institution review board). This project was approved by the ethics committee of Grenoble University (CERGA) on march 23, 2023 (CERGA-Avis-2023-08). This study was conducted according to the Declaration of Helsinki.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eKnowledge Hypergraph\u003csup\u003e\u0026nbsp;\u003c/sup\u003eapproach and main features\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eHealth data are highly connected as the link between data and information is at the basis of medical reasoning. The structure of information from CDW, based on \u003cem\u003eKnowledge Hypergraph\u003c/em\u003e [13], describes the organization of a highly connected network. The latter contains classes and hyperclasses that represent sets of \u0026ldquo;objects\u0026rdquo; and potential semantic or topological relationships between these \u0026ldquo;objects\u0026rdquo; and their \u0026ldquo;attributes\u0026rdquo;. This graphic conceptual model is flexible and suited to specify and implement the contextualization of highly connected data network elements (Figure. This method allows answering potentially complex problems through successive simple graph transversal queries [13]. First level data, or elaborated information such as medical events, can then be contextualized in relation to each other, especially temporally, even if they are structurally distant in the information structure. Moreover, \u003cem\u003eKnowledge Hypergraph\u003c/em\u003e modeling allows preserving the granularity of the original information. Finally, graphs networked according to the \u003cem\u003eKnowledge Hypergraph\u003c/em\u003e model are suitable for an interactive exploration of information, combined to build trajectories describing the longitudinal follow-up of patients [14, 16-19].\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eData extraction and elaboration\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eDe-identified data from the PREDIMED CDW included administrative and demographic information, hospitalization details, laboratory results, diagnoses, procedures, medications and free-text documents from each patient EHR. A specific data extraction from the CDW [13] was used for TRANSFUSIA, including all adult patients admitted to Grenoble Alpes University Hospital for anemia-related transfusion between 2020 and 2021. Patients were identified by crossing medical administrative data from the medical information department and medical data from the LBP registry. All patients over 18 years of age, referred to the hospital for transfusion during a short or emergency stay were considered. The first level data-set consisted of all information before and after transfusion (Supplemental material). Also recorded was all relevant information in French public epidemiology registries and administrative data. Raw data such as age, Hb level, CBC, and any numeric biological value or simple laboratory results (e.g. positive/negative), were processed to build a set of categorical variables. Categories were designed according to guidelines defining iron deficiency, i.e. ferritin level \u003cu\u003e\u0026lt;\u003c/u\u003e 30 mg/L (or \u003cu\u003e\u0026lt;\u003c/u\u003e 45 mg/L for patients over 60-year-old) and/or transferrin saturation \u003cu\u003e\u0026lt;\u003c/u\u003e 25 % [20].\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eSecond level: enrichment of the structured data network with unstructured data\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAntithrombotic treatments are often discontinued upon admission in ED, with no trace in the dataset of hospital prescription. However, doctors systematically report outpatient treatments in the initial textual medical record. An algorithm was thus built to detect any antithrombotic drug prescription in medical records of full-text ED admission reports. Finally, this new categorical variable, constituting a second level of elaborated information, was used to enrich the \u003cem\u003eKnowledge Hypergraph\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eThe final structured data network with clinically relevant information was constructed through the \u003cem\u003eKnowledge Hypergraph\u003c/em\u003e data network as an innovative approach using graph transversal queries on data organized as a network (Figure 1).\u003c/p\u003e\n\u003cp\u003eThe first event of interest was the \u0026ldquo;transfusion episode in ED\u0026rdquo;. To get a complete view of transfusion events, they were considered as a treatment of severe anemia of medical origin, However, episodes of acute anemia were not necessarily resolved in the ED, and LBP may thus have been administered both in the ED and in another hospital unit. A period of no more than 3 days was considered during which one or more LBP were administered to patients in ED. Each episode was defined by the number and type of LBP dispensed, dates of first and last administration, and their relative duration. In the \u003cem\u003eKnowledge Hypergraph\u003c/em\u003e model, the new class \u0026ldquo;transfusion episode\u0026rdquo; was created within the \u0026ldquo;treatment\u0026rdquo; hyperclass. Simple graphical queries were then used to explore the clinical context of transfusion episodes.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBased on this principle, the \u0026ldquo;emergency transfusion\u0026rdquo; event was temporally updated with the following context:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003elist of pre-transfusion laboratory tests\u0026nbsp;\u003c/li\u003e\n \u003cli\u003elist of pre-transfusion administered treatments \u0026nbsp;\u003c/li\u003e\n \u003cli\u003elist of pre-transfusion ICD10 Z51 diagnosis codes [21].\u0026nbsp;\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThe final data model is depicted in Figure 2. More detailed information is provided in Supplemental material.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eQuality control\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eBefore analyses were initiated, each set of extracted variables was screened for coherence with a pluridisciplinary check involving medical computer scientists, biostatisticians, and medical experts. Improbable values for numeric data were checked and amended or removed. Mistakes in encoding the ED stay were detected and corrected. Concerning \u0026ldquo;transfusion at ED\u0026rdquo;, there was no encoding nor typing mistake, since all LBP and whole transfusion process are electronically managed. This ensured an automatic link between the blood bank and hospital information system. All medical reports were reviewed by two hematologists to define the causes of anemia that had led to transfusion at ED. These findings were not analyzed as variables, but provided a quality check that turned out to be satisfactory. Finally, the controlled dataset was frozen before initiating statistical analyses according to local procedures following STROBE (https://www.equator-network.org/reporting-guidelines/strobe/) clinical research procedures.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eStatistics\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eCategorical variables were described as frequencies and percentages, with exact confidence intervals. Continuous variables were summarized by medians and interquartile ranges (IQR). Factors associated with early death were analyzed using Fisher exact test for categorical data. For continuous variables, comparisons were made using Student t-test for normally distributed data, or Wilcoxon-Mann-Whitney test. The normality of continuous variables was assessed through graphical inspection. All statistical analyses were conducted using R software version 3.5.2. A significance threshold of 0.05 was applied for \u003cem\u003eP-\u003c/em\u003evalues.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cem\u003eTransfusion characteristics\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eWith a starting point defined as an emergency blood transfusion episode for serious anemia in ED, evaluated within the consecutive 3 days after transfusion, 705 transfusions episodes were recorded, representing 1818 RBC units, 81 platelet units and 61 plasma units. The vast majority of transfusions required 2 LBP (median 2, IQR 2-3). Transfusions requiring 4 LBP or more were considered as related to active bleeding and secondarily relevant to intensive care unit (massive bleeding or shock) or gastro enterology department in case of digestive bleeding (Table 1).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1 Transfusion characteristics according to episodes and outcome. Univariate and multivariate analyses\u003c/strong\u003e\u003cem\u003e\u003cbr\u003e\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" align=\"\" width=\"859\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 255px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" style=\"width: 165px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eUnivariate analysis\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n=705)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"bottom\" style=\"width: 167px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMultivariate analysis* (n=671)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTransfusions N=705\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAll\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003en=705\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDeath \u0026lt; 4 months n=173\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAlive at 4 months n=532\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOdds ratio\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(95%CI)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ep-value\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOdds ratio\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(95%CI)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003eP\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e-value\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eAge years median (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e80 (66-87)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e83 (72-89)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e78 (63-86)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e1.03\u003c/p\u003e\n \u003cp\u003e(1.02-1.04)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e1.03\u003c/p\u003e\n \u003cp\u003e(1.02-1.05)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eSex (women) , n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e325 (46.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e68 (39.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e257 (48.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003cp\u003e(0.49-0.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.04\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e0.59\u003c/p\u003e\n \u003cp\u003e(0.39-0.88)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.009\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eRed blood cell consumption, Units median (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e2 (2-3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e2 (2-3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e2 (2-3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e0.87\u003c/p\u003e\n \u003cp\u003e(0.31-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e0.05\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(0.64-0.91)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.002\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003ePlatelet concentrate consumption, Units median (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0 (0-0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e0 (0-0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0 (0-0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e1.21\u003c/p\u003e\n \u003cp\u003e(0.91-1.58)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e0.17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003ePlasma concentrate consumption, Units median (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0 (0-0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e0 (0-0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0 (0-0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e1.02\u003c/p\u003e\n \u003cp\u003e(0.80-1.23)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e0.83\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eHemoglobin, \u0026nbsp;g/L\u003cem\u003e\u0026nbsp;N=672\u003c/em\u003e median (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e74.3 (16.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e75.3 (16.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e74.0 (16.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003cp\u003e(0.99-1.01)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e0.47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eLeukocytes x10\u003csup\u003e9\u003c/sup\u003e/L \u003cem\u003eN=672\u003c/em\u003e median (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e7.7\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(5.4-11.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e9.2\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(6.2-16.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e7.4 (5.3-10.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e1.05\u003c/p\u003e\n \u003cp\u003e(1.03-1.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e1.04\u003c/p\u003e\n \u003cp\u003e(1.02-1.07)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.002\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eNeutrophils, x10\u003csup\u003e9\u003c/sup\u003e/L \u003cem\u003eN=520\u0026nbsp;\u003c/em\u003emedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e5.3\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(3.7-6.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e6.4\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(3.9-11.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e5.2 (3.7-7.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e1.08\u003c/p\u003e\n \u003cp\u003e(1.04-1.12)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003ePlatelet count, x10\u003csup\u003e9\u003c/sup\u003e/L \u003cem\u003eN=669\u0026nbsp;\u003c/em\u003e median (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e227\u003c/p\u003e\n \u003cp\u003e(146-334)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e191\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(112-299)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e239 (162-342)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003cp\u003e(0.99-0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.002\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003ePlatelets \u0026lt; 50x10\u003csup\u003e9\u003c/sup\u003e/L n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e39 (5.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e17 (9.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e22 (4.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e2.53\u003c/p\u003e\n \u003cp\u003e(1.32-4.83)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.006\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eProthrombin ratio \u0026lt;70%, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e256 (36.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e94 (54.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e162 (30.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e2.72\u003c/p\u003e\n \u003cp\u003e(1.91-3.86)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e2.43\u003c/p\u003e\n \u003cp\u003e(1.64-3.61)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eINR \u0026gt;=1.2, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e326 (46.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e109 (63.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e217 (40.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e2.47\u003c/p\u003e\n \u003cp\u003e(1.74-3.52)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eAPTT\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u0026gt; 1.2, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e238 (33.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e83 (48.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e155 (29.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e2.24\u003c/p\u003e\n \u003cp\u003e(1.58-3.19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eAnti-thrombotic, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e253 (35.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e67 (38.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e186 (35.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e1.18\u003c/p\u003e\n \u003cp\u003e(0.83-1.67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e0.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eIron injection (\u0026lt;1y) n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e221 (31.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e53 (30.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e168 (31.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e0.96\u003c/p\u003e\n \u003cp\u003e(0.66-1.39)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e0.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eIron deficiency, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e263 (37.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e50 (28.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e213 (40.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003cp\u003e(0.42-0.88)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.01\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e0.51\u003c/p\u003e\n \u003cp\u003e(0.33-0.77)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.002\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eAnti-cancer treatment before transfusion in ED, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e132 (18.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e54 (31.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e78 (14.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e2.64\u003c/p\u003e\n \u003cp\u003e(1.76-3.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e2.95\u003c/p\u003e\n \u003cp\u003e(1.83-4.78)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eGI hemorrhage, \u003cem\u003eN=657\u0026nbsp;\u003c/em\u003en (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e89 (13.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e16 (9.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e73(14.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003cp\u003e(0.36-1.13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e0.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 255px;\"\u003e\n \u003cp\u003eRBC distribution index \u0026nbsp;median (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e17.2\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;(15.7-19.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e17.2\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(15.3-19.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e17.2 (15.8-20.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e0.96\u003c/p\u003e\n \u003cp\u003e(0.91-1.01)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 68px;\"\u003e\n \u003cp\u003e0.10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 87px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eCI: confidence interval; IQR: interquartile range; INR: International Normalized Ratio; APTT: activated partial thromboplastin time. ED: emergency department. GI: gastro-intestinal.\u003c/p\u003e\n\u003cp\u003e*A logistic multiple, backward regression tested independent association to early death. The model included all variables with a P-value \u0026lt; 0.1 in univariable analysis\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eCauses of anemia reviewed by experts\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eMedical record review allowed for a precise definition of the causes of anemia that led to transfusion. The main cause of transfusion in ED was confirmed to be low Hb, followed by digestive tract bleeding or liver disease (cirrhosis), hematologic disease and solid tumor. In the population presenting with early death post-transfusion, there was an over-representation of hematological malignancies and less iron deficiencies. Interestingly, the algorithmic approach identified a higher frequency of iron deficiency than experts (41 % vs 27%). However, about 60 % of anemia cases were associated to two or more causes, including iron deficiency or chronic bleeding.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eTime dimension and patient outcome\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe 705 transfusion episodes concerned 657 patients admitted at ED with anemia (emergencies relevant to trauma, cardiology, surgery or intensive care were specifically managed outside the ED and not included in the study). Taking into account the time dimension, 364 (55.4%) of these 657 patients died, 173 (26.3%) of them within 4 months after emergency transfusion. COVID positive tests at any date within the period of analysis were found in only 20 patients, representing 3 % of early deaths, yet a sensitivity analysis ruled-out any COVID effect. Table 1 compares the characteristics of the 705 episodes of patients with early post-transfusion death (n=173) and alive patients at 4 months post-transfusion (n=532), providing univariate and multivariate analyses.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eUnivariate analysis points out, in the early death population, the significance of older age and male sex. Biological parameters segregating this population are significantly higher leukocyte and neutrophil counts, lower platelet counts with more patients having less than 50x10\u003csup\u003e9\u003c/sup\u003e/L platelets, yet no difference in Hb levels. There were also significantly more altered coagulation tests in the early death population. Multivariate analysis confirmed that early death was significantly associated with age, male sex, number of LBP received, leukocytosis, and, more surprisingly, with a prothrombin ratio \u0026lt;70% and previous systemic anti-cancer therapy.\u003c/p\u003e\n\u003cp\u003eIn table 2, the main causes of anemia prompting transfusion are shown, 17% being related to a malignant condition.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2 Main causes of anemia leading to transfusion, by decreasing order\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCauses\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eN (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003eIron deficiency\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003e193 (27%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003eDigestive tract bleeding and liver disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003e154 (22%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003eChronic unknown\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003e107 (15%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003eOther\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003e83 (12%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003eHematological malignancy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003e63 (9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003eSolid tumor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003e56 (8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003ePost bleeding anemia (iatrogenic or other)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003e28 (4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003eNon-malignant hematological condition\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 302px;\"\u003e\n \u003cp\u003e21 (3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eFigure 3 provides the amount of LBP used by episode, showing the predominant use of 2 RBC bags.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn line with previous publications of prospective [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e] and retrospective [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e] studies, the \u003cem\u003eKnowledge Hypergraph\u003c/em\u003e approach used here allowed sorting out the main etiologies of anemia leading to emergency transfusion at ED, respectively iron deficiency and digestive bleeding. This study also disclosed that a higher number of LBP was transfused in ED (median 2 instead of the expected 1 as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) than in reported studies [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Of interest, TRANSFUSIA also highlighted an unexpected 25% rate of early deaths in this ED-transfused population, raising questioning about the rationale of these transfusions. Hb levels were clearly not a discriminating factor since all transfused patients had Hb levels justifying transfusion. Yet, other factors should clearly have been considered before initiating transfusion. Older patients with cancer, iron deficiency and coagulation anomalies could certainly have been taken care of by other means (IV iron transfusion, leveraged anti-coagulation) and their age and life expectancy considered before urging transfusion, with probably excessive product amounts. Since transfusions are clearly and rightly decided upon Hb levels, prior interventions to increase Hb in such nearly-terminal patients should be implemented before sending them to ED.\u003c/p\u003e\u003cp\u003eThe \u003cem\u003eKnowledge Hypergraph\u003c/em\u003e-driven data elaboration used in this study provided robust results, in a different and lighter context than that of a cumbersome and costly clinical trial. It allowed to disclose a number of ill-documented and somehow unexpected results about PBM strategies in ED.\u003c/p\u003e\u003cp\u003eThe innovative \u003cem\u003eKnowledge Hypergraph\u003c/em\u003e strategy used here could lead to new developments in the evaluation of clinical practice. Complementary to clinical studies using manually collected data, this methodology provides highly accurate audited datasets and allowed reaching an exhaustive view of transfusion at ED.\u003c/p\u003e\u003cp\u003eInitial results of the TRANSFUSIA project, currently being mainly descriptive and partially supported by the literature, are largely consistent with recent similar studies published on the subject [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. These findings confirm the robustness of the data acquisition method used here. The primary innovation of TRANSFUSIA was to use PREDIMED HDW for data collection. While randomized controlled trials remain the undisputed Gold Standard of clinical research, initial results from TRANSFUSIA offer a glimpse of research via CDWs. Indeed, they constitute a reliable alternative to traditional methods for non-interventional retrospective and prospective cohort projects, provided that the retrieved data is linked to the in-hospital records of recruited patients. However, further studies are needed to validate these initial findings. The \u003cem\u003eKnowledge Hypergraph\u003c/em\u003e approach provided a multidimensional view in terms of context of each transfusion episode that preserved logical links with patient and drugs, yet allowing for the addition of algorithmic research from non-structured information (text in medical records). Nonetheless, it took analysis a step further by adding relative time information between all elements. This is a foundation encouraging the application of further neuronal approaches in order to make predictive requirements or appropriate use of blood products.\u003c/p\u003e\u003cp\u003eBasic data such as Hb, date of transfusion, blood products, drug administration and medical reports were here exploited layer by layer, without any degradation. At the more complex layers, there was only an addition of links between primary data, and of time-related variables. This allowed considering variability with time, required for full data exploitation. Altogether, the basic data was preserved and exploited taking into account dynamics and contextualization, thus bringing more confidence in the events studied [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eMoreover, the advantage of HDWs lies in their ability to frame validation processes for data derived from multiple sources or over long time periods, a process that would require significant human resources using traditional data collection methods.\u003c/p\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003eLimitations\u003c/h2\u003e\u003cp\u003eA usual limitation of studies using CDWs lies in the current question surrounding real world health data, namely quality [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. As data from CDWs is initially collected for patient care, this has already been identified as one of the areas of improvement in CDWs, in close relationship with the notion of \u0026ldquo;fitness for use\u0026rdquo; [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. These constraints are theoretically less of an issue in prospective studies but are inherent to retrospective cohort studies [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. However, this limitation had little impact in the model applied here, considering the innovative \u003cem\u003eKnowledge Hypergraph\u003c/em\u003e approach that allows building robust contextualized information from raw data.\u003c/p\u003e\u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eBy combining a \u003cem\u003eKnowledge Hypergraph\u003c/em\u003e-driven data elaboration method and statistical analysis, the innovative approach used here provided new results for the understanding of transfusion at ED and some elements for patient management before this decision. It also provided a way of thinking about ethical views of transfusion, considering the high proportion of patients dying within 4 months after a transfusion at ED.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cul type=\"disc\"\u003e\n \u003cli\u003eEthics approval and consent to participate\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThis project was approved by the ethics committee (IRB) of Grenoble University (CERG) and was conducted in compliance with the Declaration of Helsinki.\u003c/p\u003e\n\u003cul type=\"disc\"\u003e\n \u003cli\u003eConsent for publication\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003ePatients were informed of the study and none opposed to the research (as per French law)\u003c/p\u003e\n\u003cul type=\"disc\"\u003e\n \u003cli\u003eAvailability of data and materials\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eData from this study can be made available upon reasonable request from the corresponding author after authorization from Grenoble University Hospital.\u003c/p\u003e\n\u003cul type=\"disc\"\u003e\n \u003cli\u003eCompeting interests\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThe authors declare no conflict of interest related to this work.\u003c/p\u003e\n\u003cul type=\"disc\"\u003e\n \u003cli\u003eFunding\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThis work was supported in part by institutional funding from Pfizer, Novartis, the Soci\u0026eacute;t\u0026eacute; Francophone de Transfusion Sanguine, ART.\u003c/p\u003e\n\u003cul type=\"disc\"\u003e\n \u003cli\u003eAuthors\u0026apos; contributions\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eFG designed the study, EV and FG performed extractions, CC, PAB, GD and FG developed the software, DS, AV and JLB performed statistical analyses, FG, DV, SDT, AMG and JLB wrote and edited the manuscript.\u003c/p\u003e\n\u003cul type=\"disc\"\u003e\n \u003cli\u003eAcknowledgements\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eMedical writing for this manuscript was assisted by MPIYP (MC B\u0026eacute;n\u0026eacute;), Paris, France.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eCarson JL, Stanworth SJ, Guyatt G, Valentine S, Dennis J, Bakhtary S, et al. Red blood cell transfusion: 2023 AABB International Guidelines. JAMA. 2023;3301892-1902.\u003c/li\u003e\n \u003cli\u003eStaples S, Evans H, Caulfield J, Bend M, Foy R, Murphy MF, et al. Opportunities to improve feedback to reduce blood component wastage: Results of a national scheme evaluation. Transfusion. 2024;641223-32.\u003c/li\u003e\n \u003cli\u003eLee TC, Almeida N, Pelletier P, McDonald EG. Comparison of blood transfusion rates before and after implementation of a quality improvement initiative for transfusion safety and appropriateness. JAMA Netw Open. 2023;6:e2252253.\u003c/li\u003e\n \u003cli\u003eStanworth SJ, Walwyn R, Grant-Casey J, Hartley S, Moreau L, Lorencatto F, et al. Effectiveness of enhanced performance feedback on appropriate use of blood transfusions: a comparison of 2 cluster randomized trials. JAMA Netw Open; 2022;5:e220364.\u003c/li\u003e\n \u003cli\u003eMueller MM, Van Remoortel H, Meybohm P, Aranko K, Aubron C, Burger R, et al. Patient blood management: recommendations from the 102018 Frankfurt consensus conference. JAMA. 2019;321:983-97.\u003c/li\u003e\n \u003cli\u003eMoras E, Abbott JD, Vallabhajosyula S. AABB recommends restrictive RBC transfusions for hospitalized adults and children. Ann Intern Med. 2024;177:JC14.\u003c/li\u003e\n \u003cli\u003eRadford M, Estcourt LJ, Sirotich E, Pitre T, Britto J, Watson M, et al. Restrictive versus liberal red blood cell transfusion strategies for people with haematological malignancies treated with intensive chemotherapy or radiotherapy, or both, with or without haematopoietic stem cell support. Cochrane Database Syst Rev. 2024; 5:CD011305.\u003c/li\u003e\n \u003cli\u003ePeyrony O, Gamelon D, Brune R, Chauvin A, Ghazali DA, Yordanov Y, et al. Red blood cell transfusion in the emergency department: an observational cross-sectional multicenter study. J Clin Med. 2021;10:2475.\u003c/li\u003e\n \u003cli\u003ePaganini M, Rigon F, Rebustello F, Cianci V, Bertozzi I, Randi ML. Appropriateness of packed red blood cells transfusions in chronic anemic patients in the emergency department: the TRANSFUS-ED retrospective analysis. Intern Emerg Med. 2023;18:1815-21.\u003c/li\u003e\n \u003cli\u003eBouget J, Balusson F, Viglino D, Roy PM, Lacut K, Pavageau L, et al. Major bleeding risk and mortality associated with antiplatelet drugs in real-world clinical practice. A prospective cohort study. PLoS One. 2020;15:e0237022.\u003c/li\u003e\n \u003cli\u003eGreen L, Tan J, Morris JK, Alikhan R, Curry N, Everington T, et al. A three-year prospective study of the presentation and clinical outcomes of major bleeding episodes associated with oral anticoagulant use in the UK (ORANGE study). Haematologica. 2018;103:738-45.\u003c/li\u003e\n \u003cli\u003eRaman MRG, Somu N, Kirthivasan K, Sriram VSS. A Hypergraph and arithmetic residue-based probabilistic neural network for classification in intrusion detection systems. Neural Netw. 2017;92:89-97.\u003c/li\u003e\n \u003cli\u003eCanc\u0026eacute; C, Lenne C, Artemova S, Mossuz P, Moreau-Gaudry A. Hypergraph based data model for complex health data exploration and its implementation in PREDIMED clinical data warehouse. Stud Health Technol Inform. 2022;290:335-339.\u003c/li\u003e\n \u003cli\u003eDai, Q., Gao, Y. Hypergraph computation for medical and biological applications. In: Hypergraph Computation. Artificial Intelligence: Foundations, Theory, and Algorithms. Springer, Singapore, 2023.\u003c/li\u003e\n \u003cli\u003eBoyer L, Fond G, Gauci MO, Boussat B. Regulation of medical research in France: Striking the balance between requirements and complexity. Rev Epidemiol Sante Publique. 2023;71:102126\u003c/li\u003e\n \u003cli\u003eFatemi B, Taslakian P, Vasquez D, Poole D. Knowledge hypergraphs: prediction beyond binary relations. 2020. Twenty-Ninth International Joint Conference on Artificial Intelligence and Seventeenth Pacific Rim International Conference on Artificial Intelligence {IJCAI-PRICAI-20. DOI 0.24963/ijcai.2020/299. Last accessed September 29, 2025.\u003c/li\u003e\n \u003cli\u003eChen Z, Wang X, Wang C, Li Z. A position‑aware knowledge hypergraph model for link prediction. Data Sci Engin. 2023;8:135-45.\u003c/li\u003e\n \u003cli\u003eLamine SBAB, Radaoui M, Zghal HB.\u0026nbsp;Knowledge hypergraph-based multidimensional analysis for natural language queries: application to medical data. computational science \u0026ndash; ICCS 2023: 23rd International Conference, Prague, Czech Republic, July 3\u0026ndash;5, 2023, Proceedings, Part III,\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eAydin, Berkay \u0026amp; Angryk, Rafal. (2018). Modeling Spatiotemporal Relationships Among Trajectories. 10.1007/978-3-319-99873-2_3.\u003c/li\u003e\n \u003cli\u003eIolascon A, Andolfo I, Russo R, Sanchez M, Busti F, Swinkels D,et al. Recommendations for diagnosis, treatment, and prevention of iron deficiency and iron deficiency anemia. Hemasphere. 2024;8:e108.\u003c/li\u003e\n \u003cli\u003eBr\u0026auml;mer GR. International statistical classification of diseases and related health problems. Tenth revision. World Health Stat Q. 1988;4132-36.\u003c/li\u003e\n \u003cli\u003eMaynard S, Farrington J, Alimam S, et al. Machine learning in transfusion medicine: A scoping review. Transfusion 2024; 64:162-184\u003c/li\u003e\n \u003cli\u003eEvans HG, Murphy MF, Foy R, Evans H, Li K, Wong WK, et al. Harnessing the potential of data-driven strategies to optimise transfusion practice. Br J Haematol.2024;204:74-85.\u003c/li\u003e\n \u003cli\u003eBocquet F, Campone M, Cuggia M. The challenges of implementing comprehensive clinical data warehouses in hospitals. Int J Environ Res Public Health. 2022;19:7379.\u003c/li\u003e\n \u003cli\u003eWang RY, Strong DM. Beyond accuracy: what data quality means to data consumers. J ManInf Syst. 1996;14:5-33.\u003c/li\u003e\n \u003cli\u003eMatsumoto N, Moran J, Choi H, et al. KRAGEN: a knowledge graph-enhanced RAG framework for biomedical problem solving using large language models. Bioinformatics. 2024;40:btae353.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Emergency medical services, Blood component transfusion, Artificial intelligence, Anemia, Coagulation disorder","lastPublishedDoi":"10.21203/rs.3.rs-7744688/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7744688/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe proper management of blood and labile blood products (LBP) is an important public health issue. Beside surgery, traumatology and intensive care units, as well as hematological clinics, LBP can be delivered in emergency departments (ED). The circumstances and realization of these transfusions are however ill-known and difficult to compile. Yet such information could help improve LBP management. The aim of the TRANSFUSIA study, presented here was to develop and implement an artificial intelligence approach, interpretable by medical experts, for collecting and analyzing all information relative to transfusions in EDs in a single center.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn order to evaluate this activity, a Knowledge Hypergraph was designed to synthesize the meaning of facts and numerous relationships of interest inside the datawarehouse. Its purpose was to clearly define the complex graph of structured information, organized into a network used to contextualize and produce support information of interest for clinical decision-making.\u003c/p\u003e\n\u003cp\u003eClinical and transfusion information from the years 2020 and 2021 was gathered and computed into contextualized variables of interest. TRANSFUSIA thus collected 705 transfusions from 657 patients in Grenoble University Hospital (France).Data generated were statistically analyzed.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe main aim was to analyze conditions that had led to transfusion in ED, evaluate the amount of LBP used, examine early mortality in the 4 following months and find elements to limit transfusion at ED.\u003c/p\u003e\n\u003cp\u003eTRANSFUSIA confirmed that chronic anemia and coagulation anomalies were frequent indications for emergency LBP administration and that 25% of the patients died early after transfusion. These early deaths were linked to previous anticancer treatment or abnormal prothrombin time ratios and were also impacted by age and sex.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThese results indicate a significant degree of over-transfusion in ED which could be alleviated by better prior monitoring and care of both low hemoglobin levels (to detect iron deficiency) and coagulation disorders. Additionally, the Knowledge Hypergraph- based method, developed for the TRANSFUSIA study, could be used to help decision-making facing possible LBP requirement for individual patients.\u003c/p\u003e","manuscriptTitle":"Can a factual approach using contextualized complex information provide clinical insights for appropriate transfusion at emergency departments?","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-30 23:26:39","doi":"10.21203/rs.3.rs-7744688/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2025-11-26T16:12:28+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-25T09:36:42+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"34389138118053569399312041642688237974","date":"2025-11-20T10:09:30+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"63114755460788021333050567140284765731","date":"2025-11-19T16:38:51+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-11-19T16:08:49+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-11-18T06:58:18+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-10-30T04:59:09+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-10-29T13:43:29+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Informatics and Decision Making","date":"2025-10-29T13:40:37+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"c35574aa-5a62-45c5-86d1-57d59b7c5cbc","owner":[],"postedDate":"November 30th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-11-30T23:26:39+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-30 23:26:39","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7744688","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7744688","identity":"rs-7744688","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}