Web registry links transfusion decisions to long-term patient outcomes

A prospective, web-based registry can record blood-transfusion decisions alongside the clinical circumstances surrounding them and then connect those records with patients’ long-term, self-reported health, according to a 16-month feasibility study conducted at Goethe University Hospital Frankfurt, Germany. Researchers enrolled 408 patients and documented 1,298 administrations of blood products and related therapies, demonstrating that the open-source platform could be integrated into routine care at a tertiary hospital. The pilot also exposed a major challenge for long-term research: only 12.5% of eligible patients completed the three-month follow-up and 15.2% completed the one-year questionnaire, leaving the framework operational but its longitudinal data incomplete.

Blood transfusion is a routine but complex medical intervention whose value depends on a patient’s condition, the product administered, its dose, the indication and the surrounding clinical context. Existing large databases connect transfusion exposures with health records and mortality data, but often lack the precise reasons for individual decisions and do not systematically capture how patients function months or years afterward. This limits efforts to assess guideline adherence, compare practice across institutions and determine whether variations in treatment are associated with meaningful differences in recovery. The new registry was designed to close that gap by collecting standardized clinical information at the point of care and using automated questionnaires to build longitudinal patient profiles over a planned period of up to 30 years.

The framework is called the Registry Study for Research of Blood Donor and Recipient Long-Term Outcomes and is built on OSSE, an open-source registry platform. It uses three electronic case-report forms covering basic demographics, transfusions and related blood therapies, and patient-reported outcome measures. Diagnoses and procedures are recorded as SNOMED CT concepts, allowing clinical events to be represented in a standardized, reusable format. Medical staff can enter records through a graphical interface from any internet-connected device, while participating hospitals retain equal roles in the project’s scientific governance. A centralized structure makes network-scale analysis possible in principle, although that capability was not evaluated in this single-center pilot.

Privacy is built into the architecture rather than added after collection. The OSSE system physically separates identifying information, stored in a distinct data environment, from medical information, while double pseudonymization via the Mainzelliste supports secure data exchange. Participants consent to the registry and may withdraw at any time. The approach follows data-minimization principles by limiting identifying data to fields such as name, year of birth, city of residence and contact email. The follow-up system sends a single-use personalized link by email, allowing patients to answer from any device without creating a conventional password or complex authorization process. The pilot operated in German and English, and the platform has since expanded to 16 European and Asian languages.

Clinical forms capture weight, height, documented sex, primary and secondary diagnoses, physical capacity, frailty, blood-group antigens, laboratory measurements and detailed information about each treatment. The registry records the product, date, dose, specific indication, relevant laboratory values before and after administration, and details of the clinical intervention. Physical capacity is graded in five activity-anchored metabolic-equivalent categories based on the Duke Activity Status Index, while frailty is assessed with the Clinical Frailty Scale. The electronic form also links hemorrhage risk to the anesthesiologic assessment and records discharge information. In-hospital mortality was omitted from the original pilot form but has since been added to the current iteration.

Long-term questionnaires ask patients about new diagnoses, anemia and its treatment, subsequent blood-product exposure, related adverse events, pregnancies and births, and changes in physical capacity and frailty. Every question permits an “I do not know” or “I do not want to answer” option, and the interface uses layered filters to make each questionnaire take approximately five minutes. Follow-up is anchored to the first administration, with a first questionnaire after three months and annual assessments thereafter. By the analysis cutoff of 13 August 2026, 43 of 345 eligible patients had responded at three months and 47 of 310 at one year. Paired baseline and follow-up scores were available for small subsets: 20 patients for frailty at three months, 21 for physical capacity at three months, and 20 and 14, respectively, at one year.

The enrolled population was concentrated in cardiac surgery, which limits the pilot’s generalizability. Most participants were scheduled for surgery, and 89.9% of those with a documented risk category were classified as having high expected hemorrhage risk, defined as at least 500 ml. Among 343 patients with documented sex, 245 were male and 98 female. Red-cell concentrates were the most frequently administered therapy, with 363 administrations, followed by 280 albumin administrations and 271 prothrombin complex concentrate administrations; no oral iron administrations were recorded. The registry recorded 319 distinct SNOMED CT concepts, and common entries included coronary arteriosclerosis, hypertensive disorder, aortic valve replacement and other cardiac procedures. These results show the breadth of information the platform can capture, but they describe one center’s case mix rather than transfusion practice in general.

Detailed red-cell data illustrate how the system connects decisions, biology and treatment effects. Among 363 administrations, 360 had a documented indication. The most common was a hemoglobin concentration of 7–8 g/dl accompanied by signs of anemic hypoxia or cardiovascular risk factors, recorded in 250 cases, or 69.4% of administrations with a documented indication. Pre-transfusion hemoglobin averaged 7.72 ± 0.89 g/dl in males and 7.64 ± 0.88 g/dl in females. Among 329 paired laboratory measurements, hemoglobin increased from 7.69 ± 0.89 g/dl before transfusion to 8.80 ± 0.99 g/dl afterward, a mean rise of 1.11 ± 0.99 g/dl. This before-and-after measurement documents the immediate physiological response, but the study did not perform a patient-level assessment of whether each transfusion complied with guidelines.

Sex-disaggregated analyses identified differences that the registry is intended to help investigate. A higher proportion of female patients received red cells, whereas male patients more frequently received platelet concentrate, prothrombin complex concentrate and fibrinogen. Some differences coincided with the surgical procedures recorded in the cohort: aortic valve replacement and bypass grafting were more frequent among men, while mitral valve procedures were more frequent among women. Once dosage was adjusted for body size, the investigators found no clear sex difference. The patterns therefore cannot be interpreted as evidence of inappropriate treatment, and the authors noted that comorbidity, surgical type and biometric factors require examination in larger multicenter cohorts.

Several capabilities remain aspirational. The framework is designed to support risk-adjusted institutional benchmarking, patient-level profiles, network analyses, donor outcomes and real-time dashboards, but the pilot evaluated descriptive institutional analyses only. Structured exchange through HL7 FHIR is also planned and not yet implemented, while crossmatch-to-transfusion data needed to quantify waste from unused blood-product reservations are not recorded. The low follow-up rates create possible non-response bias; although respondents resembled the wider cohort in age and sex, other differences cannot be excluded. Consequently, the paired functional assessments demonstrate that a suitable data structure can be created, not that representative recovery trajectories have already been measured. The registry continued expanding after the pilot, reaching 891 patients and 2,882 documented administrations by 11 September 2026, with a second German center being established and participation open to additional institutions.

Subject of Research: Feasibility of a prospective registry linking transfusion practice with long-term patient-reported outcomes.

Article Title: A Web-Based Registry Framework Linking Transfusion Practice to Long-Term Patient Outcomes: Design and Single-Center Feasibility Study

Article References: Nürenberg-Goloub, E., Friedrichson, B., Nordine, M., Storf, H., Zacharowski, K., & Kloka, J. A. (2026). A Web-Based Registry Framework Linking Transfusion Practice to Long-Term Patient Outcomes: Design and Single-Center Feasibility Study. Journal of Medical Systems, 50(1), Article 153. https://doi.org/10.1007/s10916-026-02474-2

Image Credits: AI Generated

DOI: 10.1007/s10916-026-02474-2

Keywords: blood transfusion, patient registry, long-term outcomes, patient-reported measures, SNOMED CT, health data, privacy by design, digital follow-up, cardiac surgery, blood safety, Web-Based, Registry