7月17日 的 H-Module 應用程式分析

Disclaimer The H-Module is a medical supporting tool used for educational purposes of the haematological acute radiation syndrome (H-ARS) only. Before making any medical decisions based on H-Module results, clinicians specialized in hemato-oncology and experienced in H-ARS should be consulted. The Threat During radiological (e.g. terrorist attack) or nuclear events (e.g. nuclear power plant accidents or use of an improvised nuclear device) subjects will be exposed to ionizing radiation. With a delay of days or weeks after radiation, injured patients will become very sick, requiring an early hospitalization and intensive care in order to survive. The Aim Physicians require rapid guidance for early and high-throughput diagnosis and therapeutic interventions of the H-ARS. Within the first three days after exposure and prior to the onset of the disease manifestation this App allows to: (1) Identify the worried well (H0) to avoid misdirection of limited clinical resources, (2) identify individuals, who will require hospitalization and if applicable intensive care (H2-4 H-ARS), (3) Identify exposed individuals, who will develop a severe/lethal degree of the hematopoietic syndrome (H3-4 H-ARS). Depending on the changes in blood cell counts, no precise allocation to a certain H-ARS severity category can be provided. In this case, a severity range will be shown and associated likelihoods of the prediction (given as positive and negative predictive values) calculated. The Tool We focused on groups of clinical significance and used logistic regression analysis to achieve a discrimination between these groups during the first three days after exposure: 1. H0 vs H1-4, identification of unexposed individuals (H0) 2 .H0-1 vs H2-4, identification of individuals requiring hospitalization (H2-4) 3 .H0-2 vs H3-4, identification of individuals who will develop a severe/lethal degree of the H-ARS (H3-4). For each of these group comparisons we examined how well changes in lymphocytes, granulocytes and thrombocytes contributed to their discrimination and build corresponding mathematical models for each day. For days 2 and 3 we examined which blood cell counts from that same day or which combination of blood cell counts from previous days (sequential diagnosis) might provide the best model for discriminating the three binary categories examined (table 1). Depending on the day and the binary category one out of these 21 models will be activated by the App. Diagnostic and therapeutic recommendations from these models are finally aggregated following an algorithm as stated elsewhere (Majewski et al. 2020). The likelihood (positive or negative predictive value) in favor of the higher or lower binary category are reflected in percent.