Transfusion-dependent Anemia
Conditions
Keywords
Artificial Intelligence, Machine Learning, Transfusion, Complications
Brief summary
Currently, about 350000 red blood cell concentrates are produced from blood donations in Austria every year. In addition to the main effect of replacing lost blood, red blood cell concentrates also have many undesirable effects - from blood group compatibilities, which are easily avoidable due to care, to storage-related side effects, to mostly intensive care problems as a result of massive transfusions, to system-wide effects such as TRALI, TACO and TRIM. Before being administered to patients, red blood cell concentrates undergo an extensive quality assurance process in which a large number of parameters are collected. Prior to use on patients, for example, bedside tests and tests for further incompatibilities with a blood sample from the intended patient are performed. With the implementation of Patient Blood Management (PBM) in recent years, the use of red cell concentrates has become more targeted - the number of transfusions is decreasing in most developed countries. However, it is still possible to suffer transfusion-related adverse events (TRAE). Thus, active research activity to reduce these TRAEs continues to be called for. To date, however, it is not known which patients experience transfusion-related adverse events. Despite the broad measures of hemovigilance and pre-transfusion testing, it is still not possible to predict which individual patient will respond to a transfusion with a typical adverse event such as hypotension, hemolysis, renal failure, or TRALI. It seems understandable that characteristics of the patient as well as characteristics of the administered unit could play a role for this. In particular, it is conceivable that a combination of characteristics of the blood unit and characteristics of the patient could determine a complication in the course of administration. For this reason, it seems attractive to use artificial intelligence and machine learning methods to predict any complications.
Interventions
BIOLOGICALTransfusion of Allogeneic Blood
Transfusion of Allogeneic Blood
Sponsors
Kepler University Hospital
Study design
Observational model
COHORT
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Inclusion criteria
* All adult patients that received at a blood transfusion the Kepler University Hospital in the period between 2016-10-31 to 2020-08-31.
Exclusion criteria
None.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| AUROC for Classification of AKI and ARF | 2016-10-31 to 2020-08-31 | AUROC for Classification of AKI and ARF |
| AUROC for Classification of AKI | 2016-10-31 to 2020-08-31 | AUROC for Classification of AKI |
| AUROC for Classification of ARF | 2016-10-31 to 2020-08-31 | AUROC for Classification of ARF |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Confusion Matrix | 2016-10-31 to 2020-08-31 | Confusion Matrix Results: true positives, true negatives, false positive, false negatives and values calculated from these results. |
| Descriptive Statistics | 2016-10-31 to 2020-08-31 | Descriptive Statistics (age in years, gender as male/female, blood group as A, B, 0, AB, hemoglobin in g/dl, platelets in count/microliter, leukocytes in cells per liter, CRP in mg/dl, creatinine in mg/dl, GFR ml/minute, glucose in mg/dl, potassium in mmol/l, TT, aPTT in seconds, ALT U/liter, paO2 mmHg, Rhesus type as positive or negative, height in cm, weight in kg, body temperature in kg, age of blood in days, cholesterine in mg/dl, systolic blood pressure in mmHg, diastolic blood pressure in mmHg) This outcome measure will compare the individual feature (e. g. height in cm) in one group vs. the other. Significant difference will be described by p-value. |
Countries
Austria
Outcome results
None listed