Novel method for trend change detection and hypothesis generation in hemovigilance: A transfusion‐related immunomodulation and blood production changes study

Abstract Background Immunomodulatory consequences of transfusion, known as transfusion‐related immune modulation (TRIM), impact patients but are not captured by hemovigilance systems. This study's objective is to explore TRIM impacts of production changes made by the blood supplier. Methods We included all transfused and non‐transfused adult inpatients from 2002 to 2022 in Hamilton, Canada. Non‐transfused patients served as controls to identify confounding temporal trends. We captured data from the Transfusion Research Utilization Surveillance and Tracking (TRUST) database and recorded TRIM outcomes including: sepsis, respiratory failure, venous thrombosis, organ dysfunction and in‐hospital mortality using International Classification of Diseases codes, Canadian Classification of Health Interventions codes, and laboratory parameters, where applicable. The blood supplier provided data on production changes and quality control assessments. We used time series trend graphs to summarize aggregate data and the rolling window E‐Divisive with Medians to detect change. A transparent and replicable point system approach identified changes in blood production most likely to have TRIM impacts. Results A cohort of 568,991 non‐transfused and 102,446 transfused hospital admissions were included. We generated 40 time series TRIM trend graphs for transfused ( n = 35) and non‐transfused patients ( n = 5). The blood supplier independently identified 12 key product policy, collection, or production changes. Consolidation of production in Ontario and introduction of buffy coat manufacturing were identified as having high TRIM impacts for patients transfused with any blood components. Conclusion Using a novel hypothesis generating data mining design, consolidation of blood production and buffy coat manufacturing are identified as changes with possible TRIM impacts among transfused hospitalized patients.