The Advent of AI in Blood Donation and Transfusion
Written by Krista Thomas, Communications Strategist, with medical direction and expertise from Dr. Tho D. Pham, Chief Medical Officer
With the rise of several artificial intelligence (AI)/machine learning (ML) applications and uses such as ChatGPT, “deep fake” AI-generated images, and even AI song contests, we are constantly encountering new areas wherein AI has the potential impact and hopefully augment existing practices. Transfusion medicine and blood donation are not exceptions; there has been significant modeling and experimentation in these fields, with some notably encouraging uses to improve patient care.
AI’s Unique Capabilities
AI has been a rapidly growing field in healthcare. One of the primary strengths of AI, particularly in this setting, is the technology’s ability to consider a tremendous amount of data and identify patterns based on what it is fed. Specifically, AI is given all the data for which we know a specific outcome; then, through various modeling programs, it can identify patterns and correlations between these huge datasets and the outcomes that may not have been apparent to humans simply by the sheer volume and variability of information. While AI certainly cannot replace human judgment in a medical context, this ability to identify and apply patterns can serve as a useful tool to provide healthcare professionals with greater insight for their individual patients and blood donors.
Application for Transfusion: When to Give Blood
One of the primary ways physicians have experimented with AI in a clinical transfusion-related setting is in determining when individual patients should receive blood products. Generally, when a patient showing no other symptoms has hemoglobin levels below a certain point, a doctor will transfuse red cells. However, each of our bodies is different and although general guidelines may help the overall patient population, any given individual may have a slightly different threshold for transfusion.
Each patient may have years of health data in their electronic medical records, but the sheer volume and variety of information make it difficult to understand which factors influence the determination to transfuse. By using AI models, many clinical and bioinformatics researchers have been able to demonstrate its ability to predict an optimal transfusion regiment for a specific individual based on their own medical history.
Although this has been done on a research basis, moving it into the clinical realm is still some ways off. Furthermore, while it is always the doctor’s broader context that guides when transfusion is used, AI can potentially add value to these patients by providing the doctor more information with which to make that final decision.
Application for Donation: How to Communicate
On the donor side of things, AI has the potential to improve crucial recruitment efforts to help donors hear from blood centers in ways that are meaningful to the individual donor. If you’ve donated with us before, you’re likely familiar with the emails we send whenever our blood supply for local patients is low. While these emails are written by heartfelt individuals on our team (don’t worry, ChatGPT isn’t asking for your blood), many elements of the messages may impact its overall reception for each person. For example, some donors may prefer to receive calls for donation during their early morning work hours, while others may only feel they have time to schedule an appointment in the evening; some donors may be moved by the story of an appreciative patient, while others may want to know the hard facts of how many units of blood are needed immediately.
These elements are ones we consider and attempt to track for the general majority whenever we send our donor community emails. However, the sheer number of variables—everything from the age of the donor to the month of the year—can make it difficult to distill meaningful patterns of effective communications.
Though we have not yet fully embarked on this exploration, we see real potential in giving AI the opportunity to work with large quantities of deidentified donor communication data to help determine which messages are the most meaningful to each individual donor, considering the many factors that have been at play for that person in the past. While it may seem a tedious task, when you consider that just one extra donor a day has the potential to save an extra life (plus the concept that, of course, we all prefer to receive communications that resonate most with each of us individually!), we feel a deeper dive into our communications is an exciting and worthy pursuit that holds the possibility of improving our ability to support the patients we serve.
With all of this focus on technology and its incredible (and at times shocking) capabilities, it can be easy to fear obsoletion. In the context of patient care, we see instead a combining of worlds that, in the end, affirms the irreplaceability of our fellow humans–especially in our times of greatest need. While we’re thankful for the advent of technology to improve possible outcomes for our patients, we’re even more thankful for the continual support of our flesh and blood donors and healthcare providers who make sure that those in our community have the products and the plans that allow them hope for healing.