Transfusion therapy has been used since before the 1900s as a palliative treatment for various medical conditions; however, the majority of transfusion demands are associated with hematological malignancies [1–3]. While numerous studies have focused on patient-specific responses to blood components, there remains a significant gap in understanding the transfusion profiles and operational dynamics of blood banks. In Brazil, hemotherapy is based on a voluntary, non-paid model as stated by National Politics of Blood and Products [4]. Data regarding blood consumption varies by Brazilian region, with the highest demand in the southeastern and southern regions [5]. The state of Amazonas is the largest by territory, yet it relies on a single blood bank located in the capital, which is responsible for supplying blood to all 39 healthcare units in the capital city, as well as to 54 municipalities across the state and several neighboring regions.

Data from 2021 indicates that over 23,000 liters of blood were distributed in the state of Amazonas, with an average of 235 transfusions performed daily. Of these, approximately 96% took place in the capital, while only 4% occurred in other municipalities across the state [6]. These Fig. underscore the critical importance of blood donation and the need for a deeper understanding of hospital transfusion profiles and patient demand for blood components.

Several guidelines have been established to provide recommendations for the clinical use of blood, with a particular emphasis on the components. Following donation, whole blood is centrifuged and separated into individual components, allowing a single donation to yield red blood cells (RBCs), platelet concentrate (PC), fresh frozen plasma (FFP), and cryoprecipitate (Cryo). These components can also be obtained through apheresis, a technique that yields higher concentrations of specific components but involves greater operational costs. Each of these blood products has distinct therapeutic purposes and plays a critical role in patient management across a variety of clinical settings [7].

Hemovigilance refers to a set of procedures that encompass the entire blood cycle, from donor recruitment to post-transfusion patient follow-up. It is primarily aimed at monitoring and preventing transfusion-related adverse events. This process is essential for maintaining quality standards, ensuring the safety of therapeutic procedures, identifying adverse reactions, and guiding system improvements [8]. As part of hemovigilance, understanding the patterns of blood component usage over time is crucial for planning future donation campaigns and optimizing resource allocation. Despite its importance, few studies have addressed this approach, even though it is well established that blood donation rates fluctuate throughout the year. Managing large datasets and extracting meaningful insights remain significant challenges for blood bank managers, further emphasizing the need for data-informed strategies in both blood donation and usage [9–12].

This study presents an analysis of hemovigilance data collected over a five-year period (2019–2023) from the regional blood center in the state of Amazonas, Brazil.

The current study reflects the transfusion reality of a reference center for hemotherapy procedures located in the northern region of Brazil. The patient profile is predominantly onco-hematological disorders. A temporal analysis of transfusion dynamics of two specialized transfusion therapy departments can contribute to strategies and local patient blood management. Interestingly, this study identified consistent increases in Cryo, irradiated CPa, and filtered & irradiated RBC over a five-year period.

The higher utilization of irradiated components observed in the present study may be attributed to the presence of an active transfusion committee. This committee establishes measures to enhance transfusion safety, minimize the risk of transfusion-associated graft-versus-host disease, and reduce dispensing errors in critically ill onco-hematological patients [13,14].

The feasibility of predicting future blood component usage was demonstrated using a Random Forest modeling approach. The model showed strong agreement with the observed values, highlighting its potential for optimizing quality control, improving the monitoring of transfusion reactions, and guiding resource allocation and strategic planning in hospital management. Additionally, the findings emphasize the importance of seasonality in transfusion patterns, which can assist blood donation campaigns, donor screening, and overall blood surveillance.

It is important to acknowledge that the study period overlapped with the COVID-19 pandemic. While studies have reported significant disruptions in blood donation and transfusion patterns during this period, mainly due to social isolation and decreased donor availability [15,16], the data of this study suggest that the trends observed in 2019 were largely maintained. An overall increase in transfusion rates was described previously [12], however, as an effect of the pandemic, a general reduction in certain blood components was noted, particularly RBC products, possibly influenced by a drop in elective surgical procedures and broader shifts in healthcare utilization during the pandemic [15]. The transfusion profile of patients seemed not to change the tendency that was observed in the previous year [2019]. The reduction seen in components, on the other hand, could have occurred either due to natural reductions or due to COVID-19, especially in respect to RBC products.

The mid-year surge in transfusion demand aligns with seasonal trends observed in other regions, especially during holidays and school vacations, when accidents and emergency procedures tend to rise [17]. Interestingly, this pattern differs from other Brazilian hospitals, such as those in Minas Gerais, where distinct temporal trends have been reported [18]. Notably, seasonal factors also affect donor availability, as holidays and family vacations coincide with a reduction in available donors [19]. In Brazil, summer school holidays occur from late December to February, with another break in July. These periods involve family travel and even affect physician availability, which may contribute to the different healthcare service patterns observed later in the year [17].

Higher demands for Cryo, irradiated PCa, and irradiated PC products were observed in the Inpatient Department, which handles acute and emergency cases. These components are often used in patients experiencing coagulopathies, undergoing oncological treatment or bone marrow transplants, or requiring hemostatic support for surgical procedures [20–23], all of which are the profile of patients attended by this department. The use of irradiated products is particularly relevant for onco-hematological patients to prevent transfusion-associated graft-versus-host disease [24–26]. Given that this hospital is the state reference center for hematological disorders [27], it is expected that such specialized components would be more frequently used in this setting.

The profile of patients in the Inpatient Department, who often present with acute bleeding conditions, supports the higher usage of hemostatic components. Despite the limited literature on blood component usage across hospital departments, the findings of this study underscore the need to tailor transfusion strategies to patient profiles and departmental needs [24]. Additionally, the increasing incidence of leukemia in the state of Amazonas [28] may reflect improved diagnostic capabilities and greater access to treatment, contributing to the rising demand for components like Cryo and irradiated PCa, which are recommended to minimize transfusion-related adverse events [29–31]. Given that patients treated at this institution often require frequent transfusions, an observed increase in certain components is expected. However, hospitals dealing with broader patient populations may exhibit different transfusion profiles.

The application of machine learning to healthcare data management, particularly in transfusion medicine, remains limited. While we utilized logistic regression, ARIMA and random forest models, there is potential for broader implementation of predictive analytics. To our knowledge, the study by Mitterecker et al. [32] remains the only work applying machine learning techniques to predict transfusion practices in general patient populations. In contrast, the present study focuses on a longitudinal assessment of a highly transfusion-dependent cohort, emphasizing the need for future research incorporating patient characteristics and follow-up data to refine transfusion management strategies.

An increase in reported transfusion reactions beginning in 2022 likely reflects the impact of a targeted training initiative introduced that year. This program included clinical case discussions, multidisciplinary education sessions, and regular departmental meetings. Consequently, reactions that previously went unrecognized began to be correctly identified and reported. This phenomenon is well-documented in Brazil and in other countries, where underreporting is often due to limited technology and data management infrastructure [33]. Sobral et al. [34] reported that 26% of RBC transfusions in the Federal District had no associated reaction report, emphasizing an urgent need for improved hemotherapy safety and documentation systems [18,35].

Time-series analysis remains underutilized in hemotherapy worldwide. Combined with the lack of technological infrastructure to process health data, especially in remote regions, this highlights a critical need for robust blood surveillance systems [33]. The current study contributes to filling this gap by presenting five years of transfusion data from a major blood center in Brazil. Nonetheless, we acknowledge limitations, including potential underreporting and the absence of an electronic system for managing transfusion and adverse reaction data.

A limitation of this study is the lack of data regarding the full transfusion supply chain; our analysis focuses on clinical administration and does not account for prior stages such as donation, processing, storage, or distribution. However, this study primarily examined the infusion phase, which represents a critical point in transfusion practice with direct implications on patient outcomes. Our results provide a better comprehension of blood component patterns from a specialized center, to mitigate further storage protocols and implement new public health strategies and professional training.

The findings of this study demonstrate that the patient’s profile has a direct influence on transfusion demand but more importantly highlights the importance of local analysis to comprehend blood demand heterogeneity in Brazil. Future studies must integrate data from all the steps of the blood transfusion cycle to provide a broader point of view and enhance a national overview of blood usage in the country.

In recent years, many hospitals have implemented new technologies and tools aimed at identifying innovative administrative strategies with greater efficiency, guiding future public health initiatives with higher impact and improved cost-effectiveness for health systems. The application of machine learning techniques holds considerable promise for enhancing data analysis, interpretation, and decision-making across various healthcare environments, including hospital settings. However, the application of such approaches within the field of transfusion medicine remains limited, due to both intrinsic and extrinsic challenges that hinder the development of effective public health measures in this domain.

The present study identified a consistent increase in the use of irradiated platelet concentrates (irradiated PC and irradiated PCa) and Cryo over the past five years, reflecting a growing demand for these components among patients with hematological and onco-hematological conditions. Furthermore, stratification by hospital department revealed that the Inpatient Department, responsible for urgent care, exhibited a higher demand for blood components associated with hemostatic support, consistent with the underlying disease burden in this particular patient population.

The data presented in this study contribute not only to local hospital planning but also to a broader understanding of transfusion practices. Analyzing temporal trends and integrating data through advanced technological tools should be considered a priority. These processes are vital for optimizing clinical decisions and guiding management strategies in transfusion medicine.

Conceptualization: T.A.L., J.S.C., A.L.S.-J. and E.C.C.; Data curation: C.H.A.C., T.R.A., A.L.S.-J. and E.C.C.; Formal analysis: J.A.C.C. and A.L.S.-J.; Funding acquisition: A.L.S.-J. and E.C.C.; Investigation: C.H.A.C., T.R.A. and J.A.C.C.; Methodology: C.H.A.C., T.R.A. and J.A.C.C.; Project administration: T.A.L., J.S.C., A.L.S.-J. and E.C.C.; Supervision: J.S.C., A.L.S.-J. and E.C.C.; Visualization: J.A.C.C. and A.L.S.-J.; Writing – original draft: C.H.A.C., T.R.A., and J.A.C.C.; Writing – review & editing: T.A.L., J.S.C., A.L.S.-J. and E.C.C. All authors reviewed and approved the final draft of the manuscript.

Prior to data collection, this study was approved by the ethics committee of the Fundação Hospitalar de Hematologia e Hemoterapia do Amazonas (HEMOAM) under protocol number #6.575.757.

C.H.A.C. and T.R.A. holds an undergrad scholarship from Fundação de Amparo à Pesquisa do Amazonas (FAPEAM) via Programa de Apoio à Iniciação Científica from HEMOAM (PAICHEMOAM) for this work. J.A.C.C. and E.C.C. hold a MSc and PhD scholarship, respectively, from Fundação Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES). A.L.S.-J. holds a post-doctoral fellowship from Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP). Any of the funders had any decision on conceptualization, design, data collection, analysis, decision to publish nor preparation of this manuscript.

The data that support the findings of this study are available from the corresponding author upon reasonable request.