Search Results

AI tools are showing up everywhere in medicine right now — in our inboxes, in meetings, and quietly in the background as we prepare talks or look up unfamiliar territory. Many of us are experimenting with them in real time, often between consults or after a busy clinic day, trying to figure out what they’re actually good at and how to use them without creating extra work. One place where AI can be genuinely helpful is in orienting yourself to a clinical question — especially when you need a quick overview before diving deeper. Over the past year, I’ve found that pairing AI tools with traditional verification steps has made my own literature searches faster and more organized, while still keeping the process grounded in real evidence. Since many colleagues are exploring these tools too, I thought I’d share the simple workflow I’ve settled into. Nothing here is prescriptive; it’s just what I’ve found useful as a clinician who wants speed and  reliability. What AI Can Do Well (and Why It’s Helpful) AI can be a surprisingly helpful companion when you’re approaching a clinical topic. It can: Summarize large volumes of text quickly Highlight themes or connections across papers Provide a starting point when you’re approaching a topic you haven’t revisited in a while Help double-check that you’re not missing obvious papers Turn unstructured information into something more organized AI isn’t a replacement for reading source papers, but it can  make it easier to start with some structure already in place. My Three-Step Workflow 1. Start With OpenEvidence OpenEvidence has become my go-to for initial orientation. It’s built specifically for medical literature and has content agreements with NEJM and JAMA, which helps anchor it in reputable sources. What I appreciate most is that every statement comes with a citation, and you can click directly into the underlying study. Two very practical notes: It’s free for medical professionals, which makes it easy to recommend. There’s also a mobile app, which is surprisingly handy when you’re on service and need to look something up between cases. For me, OpenEvidence gives a quick landscape of what has  been studied, what hasn’t , and where the evidence feels solid versus sparse. Website: https://www.openevidence.com/ 2. Cross-Check and Structure With Elicit I don’t use Elicit for every question, but I often reach for it when I’m working on publications, talks, or anything where I need to be comprehensive. Elicit is trained on a broader scientific corpus, which means it sometimes pulls in studies that OpenEvidence misses or adds contextual pieces that help round out the picture. Its real strengths are: generating tables from search results extracting sample sizes and primary outcomes grouping related studies summarizing PDFs you upload If OpenEvidence helps me understand the landscape, Elicit helps me organize and structure that landscape — especially when multiple study designs or subtopics are in play. Website: https://elicit.com/ 3. Verify With a DOI Check (My Favorite 10-Second Step) Once I’ve identified the key papers, I take the DOI or PubMed ID and paste it directly into Mendeley, which will automatically fetch the citation metadata and abstract. A few reasons I rely on this step: It confirms the paper exists The metadata is correct The journal, year, and authors match The abstract aligns with the AI summary Not all reference managers can fetch metadata from just a DOI or PubMed ID — but Mendeley can, and Mendeley is free, which makes it a great option if you need an accessible verification tool. This small step has saved me more than once from citing a misattributed or nonexistent paper. A Gentle Note on Limitations AI tools are still evolving, and so are we. They can miss studies, overstate certainty, or conflate adjacent concepts. That’s not a failure — just a reminder that they’re best used alongside our clinical judgment and our usual habits of checking primary sources. For me, the workflow above keeps things balanced: AI helps with speed and structure, and the DOI check keeps everything grounded in reality. When This Workflow Helps Most I reach for this system when: Preparing for a meeting or protocol discussion Refreshing a topic I haven’t touched in a while Getting oriented before reading more deeply Drafting a talk, manuscript, or background section Checking whether references actually exist before citing them This workflow is flexible: I use it for everything from quick orientation to deeper literature reviews. The steps stay the same; the depth just changes depending on the question. This has become my primary approach to reviewing the literature -- it fuses speed with reliability in a way that fits how we practice today. I still use PubMed when I need to dive deeper into a particular thread, but the core workflow starts here. Closing Thoughts AI is becoming part of everyday clinical practice, and most of us are learning as we go. My hope is that sharing this workflow helps demystify the process a bit and gives you a reliable and practical starting point if you’re exploring these tools yourself. If you’ve found other strategies or tools that work well for you, I’d genuinely love to hear them — we’re all figuring this out together.

There are days in Transfusion Medicine when the most interesting teaching moments arrive quietly — between phone calls, in the apheresis unit hallway, or as someone leans back in a rolling chair and says, “Okay, but why can’t we just do a depletion exchange here?” Today it came up while troubleshooting an inpatient red cell exchange on a Sickle Cell patient who was a lot sicker than he’d been two weeks earlier. One person suggested adding a depletion phase to improve the efficiency of the run. And that’s when the conversation shifted — away from algorithms and toward physiology, which is where these decisions actually live. Because the truth is simple: Depletion exchange works beautifully — until it doesn’t. And the people for whom it can go wrong are exactly the ones who can’t afford a period of reduced oxygen delivery. What Exactly Is a Depletion Exchange? Before diving into the “why not,” it’s worth being clear about what a depletion exchange actually is — because the term gets thrown around loosely, and not everyone pictures the same thing. A depletion exchange , also known as isovolemic hemodilution red cell exchange, is a specific variant of automated RCE in which the procedure begins with a hemodilution phase. The sequence looks like this: First, patient RBCs are removed. The device takes off red cells from the circulating volume. Simultaneously, the machine replaces the removed volume with crystalloid or 5% albumin. This maintains volume  (isovolemia) but not oxygen-carrying capacity. After the patient’s hematocrit is intentionally lowered, the machine proceeds with the regular red cell exchange phase — removing patient cells and replacing them with donor RBCs. The rationale is straightforward: By lowering the patient’s starting hematocrit, each donor unit becomes more “effective” at reducing HbS%, so fewer units are needed. It increases efficiency, reduces donor exposure, and improves the geometry of the exchange. But there is a catch — and it’s the one people forget: You are creating a temporary period of reduced oxygen delivery. Isovolemic ≠ iso-oxygenating. For most stable outpatients, that’s fine. For others, it’s the wrong physiologic bet. Once you see the mechanics laid out like that — the intentional dip in hematocrit, the temporary thinning of oxygen delivery — the real issue isn’t the technology at all. It’s the patient . And there are certain patients whose physiology simply can’t afford that moment of dilution. 1. Acutely Ill Inpatients: No Physiologic Room to Fall We see this all the time: The patient with acute chest. The patient with sepsis layered on top of pain crisis. The patient who walked in hypoxic and is now teetering at 94% on oxygen. These patients are already running on borrowed reserve. Even a short-lived decrease in hematocrit can widen the gap between “holding steady” and “crashing.” Their tissues are extracting everything they can. Their compensatory mechanisms are maxed. A brief dilution phase risks exactly what we’re trying to prevent: worse perfusion, more ischemia, more instability. So we skip depletion. Not because we can’t do it, but because they can’t afford the physiologic tax. 2. Pregnancy: Two Patients, One Oxygen Supply Pregnancy is its own cardiovascular universe — high output, reduced systemic vascular resistance, compressed venous return, and a placenta that is exquisitely sensitive to maternal perfusion changes. The math is simple: Lower maternal Hct → lower uteroplacental oxygen delivery. Even transiently. Even “just during the depletion phase.” And when oxygen delivery falters, the fetus feels it first. In the interest of safety, we do not perform depletion exchanges in pregnant patients. 3. Cardiac Patients: Tightly Balanced at Baseline Then there are the patients with cardiac histories — and the patients with cardiac histories they don’t know they have yet. In cardiology, the pendulum has swung back toward liberal transfusion strategies for acute coronary syndromes, with several recent studies showing improved outcomes when hemoglobin is kept closer to 10 g/dL rather than drifting down into the restrictive ranges. The reason is simple and intuitive: the ischemic myocardium hates anemia. Coronary perfusion is already limited; oxygen extraction is already maxed. Any additional dip in oxygen-carrying capacity — even brief — can worsen supply–demand mismatch. And that’s the core problem with depletion exchange in this population. The machine keeps the volume  steady, yes — but it cannot shield the myocardium from the temporary but real drop in perfusion during the dilution phase. It’s a moment the heart has no margin to absorb. So for these patients, we choose safety. Exchange only. Slow and steady. So Why Do We Do Depletion at All? Because when it’s safe — in stable outpatients without physiologic red flags — it is  useful. It can make the exchange more efficient. It can reduce donor exposure. It can improve the final HbS% with fewer units. But the moment someone is acutely ill, pregnant, or carrying cardiac risk, those advantages don’t justify even a temporary hit to oxygen delivery. Apheresis Isn’t Just a Machine. It’s Physiology. That was really the take-home from today’s conversation. Our protocols can get so algorithmic that it’s easy to forget the body isn’t following the same neat logic tree. There’s a human being on the other side of the circuit — one who may be running out of compensatory room. So when we pick the exchange modality, we aren’t just choosing a setting on the instrument. We’re declaring what we think the patient can physiologically tolerate. For some, dilution is a gift. For others, it’s a risk not worth taking. And the art — the part that never shows up in the software — is knowing the difference.

Introduction: The Problem With the "Second Brain" Metaphor Artificial intelligence in pathology and laboratory medicine is often marketed with an irresistible promise: a second brain that will spot what humans miss, automate the tedious parts of practice, and bring order to the overwhelming volume of data moving through modern health systems. It’s a compelling metaphor—but also a deeply misleading one. The truth is simpler and far more useful: most AI tools in lab medicine today are not second brains. They are second readers. They assist. They triage. They flag patterns. They highlight outliers. They nudge clinicians toward questions worth asking. This is not a limitation—it is the sweet spot of responsible AI. The problem is that our metaphors, expectations, and sometimes our implementation strategies haven’t caught up with this reality. When we treat assistive AI as if it were autonomous, we misjudge both its power and its risks. This piece reframes AI in pathology and transfusion medicine through a more grounded, clinically realistic lens: AI as a second reader—never the primary decision-maker. Assistive vs Autonomous AI: Why the Distinction Matters In public conversations, "AI" tends to be treated as a single monolithic category. But in clinical practice, the distinction between assistive and autonomous systems is foundational. Assistive AI Assistive AI tools support human decision-making without replacing it. They: flag abnormal cells or slide regions for review, surface unusual utilization patterns, predict inventory needs, identify potential bleeding risks or outlier transfusion practices, augment quality control workflows. The human remains the final decision-maker. The AI's role is advisory. Autonomous AI Autonomous AI, by contrast, can issue a clinical interpretation without human confirmation. The classic example is FDA-cleared autonomous diabetic retinopathy screening, where the system renders a result independently. Pathology is not there—and ethically, operationally, and scientifically, it shouldn’t aspire to be. Tissue interpretation, pre-analytic variability, complex clinical context, and downstream consequences place pathology squarely in the domain of human-in-the-loop practice. Moreover, the limitations of autonomous AI make full automation particularly risky in this field. Even state‑of‑the‑art large models exhibit irreducible error rates, including hallucinations that arise not from software bugs but from the fundamental way probabilistic systems generate outputs. OpenAI and other major developers have acknowledged that hallucinations are inevitable  in current-generation AI—an acceptable risk for drafting emails, but not for diagnosing malignancy. In pathology, an autonomous error is not a benign failure mode; it is a misdiagnosis. Slides vary between institutions, stains differ, scanners introduce artifacts, and rare entities can be misclassified with absolute confidence. The model does not know when it does not know. Human-in-the-loop practice is therefore not a philosophical preference but a safety requirement. Current professional sentiment reflects this: most pathologists are cautiously optimistic about assistive AI but deeply wary of autonomous systems. The field understands that algorithms can elevate quality and efficiency, but they cannot—and should not—bear sole responsibility for interpreting tissue, integrating clinical nuance, or adjudicating uncertainty. Why the distinction matters Marketing narratives blur the line between assistive and autonomous. Operationally, this creates two dangerous extremes: over-trust: assuming the model "knows" more than it does, under-trust: dismissing or ignoring helpful signals because expectations were unreasonable. Treating AI as a second reader  helps calibrate our expectations and clarifies the respective responsibilities of humans and machines. Workflow, Not Math: The Hidden Barriers to Clinical Integration Technical performance is rarely the limiting factor for AI deployment in lab medicine. More often, the barriers are operational and workflow-driven. Pre-analytic variability No algorithm, however elegant, can overcome poor input. Hemolysis, mislabeled samples, incomplete clinical information, and inconsistent sample handling all degrade model performance. "Garbage in, garbage out" is not cynicism; it is clinical reality. LIS/EMR integration An AI flag that never reaches the transfusion physician or technologist in a usable format is functionally irrelevant. Many promising tools fail not because they are inaccurate, but because they exist outside the everyday workflow. Alert fatigue If an AI model surfaces insights the same way EMR pop-ups surface medication alerts, clinicians will click through them reflexively. Effective AI must blend into the workflow — not interrupt it. Staff training AI disagreement is a liminal space. When a model flags an unexpected pattern, what is the technologist supposed to do? Without clear protocols, the burden on staff increases rather than decreases. Model stewardship Who revalidates the model yearly? Who monitors drift? Who owns threshold adjustments? Governance is critical and cannot be an afterthought. These challenges are not exciting, but they are what determine whether an AI tool genuinely helps clinicians — or becomes abandoned. The Hype Cycle Problem AI in medicine moves in predictable hype cycles. When expectations are unrealistic, three harms follow: 1. Overpromising leads to disillusionment When leadership expects instant automation, disappointment is inevitable. This can poison the well for future tools that are more modest but more practical. 2. Steps get skipped Proper change management, validation, and staff training take time. Under the pressure of hype, institutions try to "roll out" tools before anyone understands how to use them. 3. Trust becomes polarized Some clinicians embrace AI uncritically. Others reject it entirely. Neither posture produces safe patient care. Reframing AI as a second reader  helps temper the hype and brings expectations back into alignment with clinical workflow and real-world constraints. What Safe, Responsible AI Actually Looks Like Clear intended use Every AI tool must answer one question precisely: What is the intended use? Ambiguous purpose leads to ambiguous outcomes. Human-in-the-loop structure High-impact clinical decisions — transfusion thresholds, rejection of critical values, or product allocation — should never be automated fully. AI highlights patterns; humans interpret them. Local validation Models must be calibrated to local population characteristics, including major demographic differences, high-obesity populations, rare disease prevalence, and unique practice patterns. Ongoing monitoring Performance changes over time. Drift is real. Monitoring is not optional. Defined failure modes Clinicians need clarity: When should I ignore this model?  Understanding limits is as important as understanding utility. Explainability (pragmatic, not academic) Technologists and clinicians need broad insight into why a model fires — high-level logic is sufficient. Full algorithmic transparency is not required. Together, these guardrails ensure that AI functions as a clinically meaningful assistant, not an unpredictable black box. A Transfusion Medicine Lens: Where AI Actually Delivers Value Transfusion medicine offers a prime example of how AI should function in practice: as a second reader that enhances safety and efficiency. Utilization and stewardship AI can identify patterns of overuse or underuse, highlight outlier ordering habits, or flag cases where restrictive thresholds are inconsistently applied. But humans — transfusion physicians, technologists, PBM programs — interpret and respond to these patterns. Inventory and product management Platelet forecasting, rare phenotype prediction, and resource allocation are well-suited to assistive AI. The model surfaces the signal; the human makes the plan. Risk prediction Predictive models for bleeding, DHTR risk, TRALI likelihood, or massive transfusion activation can bring subtle risk factors to the surface. They augment human judgment but do not replace it. These examples demonstrate the core argument of this piece: AI helps most when it supports human cognition without competing with it. Conclusion: Getting the Metaphor Right AI in pathology and laboratory medicine is not a second brain—and expecting it to be one sets everyone up for failure. It is a second reader. A pattern spotter. A triage assistant. A flagger of outliers. A partner in safety and quality. When we ground AI in its true purpose, we can finally deploy it in ways that are meaningful, safe, and sustainable. The challenge is not to automate pathology or transfusion medicine, but to integrate AI into workflows as a thoughtful collaborator. The future of AI in the laboratory will belong to the institutions and clinicians that understand this distinction: Useful AI is not autonomous. It is assistive — and that is exactly where it belongs.

I’ve been thinking a lot about attribution bias lately — the reflex to explain someone’s behavior by pointing to their character instead of their circumstances. In medicine, this isn’t just a cognitive shortcut. It’s one of our favorite misdiagnoses, and it often shows up in the form of a single, damning label: the difficult patient. Two encounters from my own practice keep coming back to me. 1. “The Meanest Person I’ve Ever Met.” That was the handoff. The wife was “the meanest person I’ve ever met.” “Confrontational.” “Always angry.” “Impossible to deal with.” This is a common setup: a pre-labeled human wrapped in warning tape, delivered with the expectation that I will treat her like a hazard. But her husband was lying in an ICU bed because of a botched procedure that left him paralyzed. She was navigating trauma, grief, and a system that — because of medicolegal anxieties — had decided to keep her at arm’s length and speak around her instead of to her. Every door she knocked on had a sign that said You may enter; we will not tell you anything of substance. When I met her, I didn’t find the “meanest person.” I found a woman trying to save what was left of her life. She wasn’t hostile. She was frantic. She wasn’t aggressive. She was afraid. She wasn’t difficult. She was drowning. Nothing about her behavior was surprising once you considered the situation. 2. “Behavioral Issues” in an Incarcerated Patient The second case came wrapped in a different set of labels: “behavioral issues,” “noncompliant,” “gets angry,” “hard to talk to.” An incarcerated Black man with sickle cell disease — a combination that, in the hospital, often guarantees dehumanization from the start. He’d been spoken to over the shoulder, not face-to-face. Guards in the doorway. Clinicians darting in and out, clipboards between them and him. A patient assessed through a frame of suspicion before a single word was exchanged. So I did something radical in its simplicity: I sat down. I looked him in the eyes. I spent twenty minutes listening. He was delightful. Honest. Funny. Thoughtful. A person. The “behavioral issues” vanished the moment the assumptions did. The Failure Isn’t the Patient. It’s the Attribution. This is attribution bias at its most damaging: mistaking trauma for personality, mistaking fear for hostility, mistaking systemic failure for individual flaw. In medicine, we love tidy trait-based stories — she’s mean, he’s manipulative, they’re noncompliant — because traits feel permanent. Predictable. Containable. But traits are the least accurate predictors of behavior, especially in crisis. Most human behavior is not driven by character. It is driven by emotional state. It ’s driven by: fear pain powerlessness being unheard being dismissed being rushed being judged being visibly feared feeling unsafe All of these are situational. All of them are correctable. None of them are personality traits. Fixed Mindset Medicine vs. Growth Mindset Humanity Attribution bias is rooted in a fixed mindset: the belief that people behave the way they do because of unchangeable internal qualities. But humans are not static. Neuroscience makes this painfully clear. Our prefrontal cortex — the part that lets us reason, regulate, pause, and plan — is resource-hungry and fragile. When people are in crisis, their frontal lobes go offline and their limbic systems take the wheel. They become emotional, reactive, short-fused, protective. Not because they’re “bad” or “difficult,” but because they’re human. In other words: Behavior = situation × current state × available resources (not “behavior = personality”). A growth mindset — the belief that behavior is modifiable and context-dependent — is not a soft, feel-good philosophy. It’s a neuroscientific reality. It also makes us better clinicians. The Stories We Tell Shape the Medicine We Practice Once we label someone as “difficult,” we stop asking the essential questions: What happened to them? What are they afraid of? What do they need to feel safe? What system-level failures are shaping this interaction? And maybe the hardest one: Who would I like to be in their situation? That question alone could dismantle half of the attribution bias in our hospitals. The Truth Behind Most “Difficult Patients” If I’ve learned anything, it’s this: Patients are almost never difficult because of who they are. They are difficult because of what they’re going through —and because of how the system is treating them. Change the situation, and the behavior changes. Change the framing, and the person emerges. Change how we show up, and the whole encounter transforms. We don’t have “difficult patients.” We have difficult circumstances — and patients doing their best within them.

Two different patients. Two plasma exchange treatments. Two nurses asking me, gently and matter-of-factly, the same question: “Do you want to chase with some plasma?” Before becoming an attending, I had never heard the phrase. It wasn’t part of residency, fellowship, ASFA courses, or any protocol I’d ever followed. It certainly isn’t in textbooks. The first time someone asked me, I wondered whether this was a regional term or a long-standing tradition I’d somehow missed. What I’ve realized is that “plasma chasers” aren’t a formal practice at all—they’re a local solution to a real physiologic concern, passed down through experience rather than evidence. And once I understood that, the whole thing made much more sense. Two Encounters, Two Decisions The first patient had an endomyocardial biopsy two days before their TPE. That felt like a clear “yes.” Even a small pericardial bleed can turn into tamponade quickly, and albumin-only exchanges temporarily lower fibrinogen in exactly the wrong moment. The second patient had a chest-tube exchange two days prior. Compressible, external, and not associated with catastrophic rebleeding after 48 hours. That one was a comfortable “no.” Both decisions felt reasonable. But afterward, I found myself thinking about why the question exists in the first place — and why different centers use different rituals to manage uncertainty. Ambiguity Breeds Ritual During my PhD years, I saw how easily small rituals form in the lab. The postdoc who said you had to swirl counter-clockwise for best DNA yields. The technician who swore PCR only worked if she spun down tubes twice. The graduate student who insisted cells behaved better if passaged on Tuesdays. None of these traditions were harmful. They were simply the human response to complex systems with hidden variables. When outcomes are unpredictable and stakes feel high, it’s natural to reach for anything that offers a sense of control. Clinical medicine is no different. Apheresis has many moving parts, physiology we can’t always observe directly, and very little high-quality evidence for the fine details of practice. It’s not surprising that different institutions develop their own habits — some sound, some questionable, some simply inherited. “Plasma chasers” live right in that space. What the Survey Data Actually Tell Us Before I wrote this, I went looking for anything peer-reviewed about plasma chasers specifically. There isn’t anything — not a single survey or guideline entry. But there is  a published ASFA-linked survey (Zantek et al., J Clin Apher 2018) about hemostasis management and replacement fluid decisions. And the results were eye-opening: When a patient had major surgery just one day earlier, 8.9% of respondents still used albumin-only  replacement, a much higher percentage than I expected. For minor procedures one day prior, 49.5% used albumin-only, and 50.5% included some or all plasma. That’s about a 50/50 split. For a patient scenario with no bleeding risk, 94.7% used albumin-only. Which is still short of 100% like I expected. To me, that’s fascinating. It shows how inconsistent — and how intuitive — these decisions really are. Clinicians are already making judgment calls about post-procedure bleeding risk every day, even without formal algorithms. Plasma chasers are simply a more granular version of that same instinct: Does this patient need some factors right now? Could a small bleed matter? The Framework That Actually Makes Sense When I strip away the inherited rituals, peer pressure, institutional memory, and “this is how we do it here,” the physiologic picture becomes surprisingly straightforward. Use a plasma chaser when: The patient had an endomyocardial biopsy < 72 hours. The patient had a renal biopsy < 72 hours. There is a fresh injury in a space where even a small bleed can be dangerous before it becomes obvious. These are the scenarios where a little post-exchange factor support truly makes sense. Consider partial FFP replacement when: A patient has severe allergic reactions to plasma but still needs some factor replacement. A patient is highly citrate-sensitive, and full FFP carries risks. Use full FFP replacement when: The indication is TTP. There is active or recent major bleeding. There is a high-risk coagulopathy. Use albumin-only when: A small bleed won’t be catastrophic, such as with chest tubes, lumbar punctures, and GI biopsies. There’s no compelling reason for factor support. This framework isn’t mystical. It isn’t ritualistic. It’s just physiology, risk, and common sense. Why I’m Writing About This I’m not criticizing the practice of plasma chasers. In many ways, I admire the quiet wisdom embedded in these unofficial patterns of care. They represent clinicians trying to do the safest thing for their patients in a landscape where evidence is incomplete. But I also believe there’s value in naming the uncertainty, reflecting on it, and disentangling ritual from reasoning. I don’t think we talk enough about the gray zones in our specialty — the places where we make decisions based on physiology, pattern recognition, and a little bit of fear of the worst-case scenario. And I think there’s a kind of comfort in acknowledging that these instincts come from somewhere real. Because in the end, apheresis is full of places where the science is incomplete, and the art of medicine steps in — not as hoodoo, but as thoughtful, experience-guided care.

For years, transfusion-associated circulatory overload (TACO) has been framed as a purely hemodynamic problem — a case of too much blood, too fast.  But hemovigilance data are challenging that simplicity. A growing body of work suggests that in a significant subset of patients, TACO runs hot. Yes, fever. Not chills from contamination, not cytokine-release fever from a leukocyte-rich product, but true fever within hours of transfusion — sometimes the only obvious clue that something is wrong. And it’s not rare: recent studies suggest that 30–40 % of TACO cases involve fever, a rate higher than for allergic transfusion reactions with fever. [1- 3] Beyond Volume: A Hotter Kind of Overload Classically, TACO is defined by acute respiratory distress and hydrostatic pulmonary edema within six to twelve hours after transfusion. But the presence of fever doesn’t fit that simple model of mechanical overload. Research by Parmar et al. (2017) and others shows that these fevers aren’t linked to patient age, product age, or reaction severity — and in most cases they’re new-onset, not continuations of pre-existing fever. [1] Together with bedside biovigilance data showing inflammatory features in some TACO cases, [2 - 4] this has led to a re-imagining of the syndrome: TACO may be part hemodynamic, part inflammatory. The Two-Hit Hypothesis: Volume Meets Inflammation The two-hit hypothesis of “inflammatory TACO” frames the reaction as a meeting of two vulnerabilities: First hit: a susceptible patient — one with heart failure, renal disease, positive fluid balance, or critical illness that limits their ability to tolerate volume. Second hit: the transfusion itself, delivering not only volume but also biologically active mediators — cytokines, storage-lesion byproducts, and shifts in colloid osmotic pressure. This combination may tip the endothelium into dysfunction, increasing capillary permeability and producing pulmonary edema beyond what simple volume overload would explain. It also helps account for “hot-TACO” cases after even a single unit of blood. [2 - 4] Clinical Confusion: When “Hot-TACO” Mimics TRALI Fever blurs the lines. In a febrile, hypoxic patient post-transfusion, most clinicians first suspect TRALI or sepsis. Yet as multiple studies and the revised international case definition emphasize, [1, 3, 5] the presence of fever doesn’t exclude TACO. If there are clear hydrostatic findings — positive fluid balance, elevated BNP or NT-proBNP, echocardiographic evidence of elevated filling pressures, or improvement with diuretics — TACO should remain high on the list even when fever is present. Diagnostic Pearls: Sorting the Hot from the Heavy When TACO and TRALI overlap, these clues help steer the differential: 🕒 Timing: TACO usually develops within 6 hours, but may be delayed up to 12. TRALI is classically within 6 hours and not relieved by diuretics. 💧 Volume response: Improvement with diuretics or fluid restriction supports TACO. ❤️ BNP / NT-proBNP: Ratios > 1.5–2× pre-transfusion favor hydrostatic overload. 🫁 Chest imaging: TACO shows cardiomegaly and vascular redistribution ; TRALI typically presents with bilateral non-cardiogenic infiltrates. 🧪 Inflammatory markers: Fever alone doesn’t rule out TACO, but a marked cytokine surge (e.g., IL-8, IL-6) suggests TRALI or sepsis. Ultimately, distinguishing hot-TACO from other febrile transfusion reactions depends on pattern recognition rather than a single test. The key is to remember that not all TACO is “cold.” Sometimes, the circuit overload burns a little. References Parmar N et al. Vox Sanguinis.  2017;112(1):70-78. Andrzejewski C et al. Transfusion.  2012;52(11):2310-20. Wiersum-Osselton JC et al. Lancet Haematology.  2019;6(7):e350-e358. Bulle EB et al. Blood Reviews.  2022;52:100891. Delaney M et al. Lancet.  2016;388(10061):2825-2836.

A patient was admitted with a congestive heart failure exacerbation. Their hemoglobin was drifting downward — nothing dramatic, but enough to warrant a type and screen. The result wasn’t surprising: a known warm autoantibody. What was  surprising was the note that popped up beside it — “Requires washed RBCs.” We looked into it. The patient’s IgA level was reported as < 5 mg/dL on two separate occasions — a true, complete selective IgA deficiency. No history of anaphylactic reactions, no documentation of transfusion reactions at all. Still, the washed requirement persisted, a permanent flag carried forward through admissions like a family heirloom no one quite questioned. The Spectrum of IgA Deficiency Selective IgA deficiency is the most common primary immunodeficiency, occurring in roughly 1 in 300 people, though the term encompasses a spectrum. Many individuals have low but detectable  levels of IgA and remain entirely asymptomatic. A complete  deficiency — defined by an undetectable IgA level on at least two separate occasions — is far less common. (This definition is used by the European Society for Immunodeficiencies and the Immune Deficiency Foundation.) Only a fraction of these individuals go on to form anti-IgA antibodies, which have been implicated in allergic or anaphylactic transfusion reactions. The Rare Meets the Real The classic teaching looms large in every pathology and transfusion board prep book: the IgA-deficient patient who develops life-threatening anaphylaxis after receiving a standard blood component. But outside the exam room, this scenario is exceedingly  rare.The true incidence of anti-IgA–mediated anaphylaxis is unknown and appears extremely low. The literature contains only a handful of case reports and small series describing such reactions, mostly in patients with severe  IgA deficiency and detectable  anti-IgA antibodies [1–4]. A comprehensive review identified just 23 cases of anaphylaxis in immunodeficient patients receiving IVIG over several decades [2]. Even among those with measurable anti-IgA, many tolerate blood products and immunoglobulin infusions without incident [1]. The association between anti-IgA antibodies and anaphylaxis remains controversial — suggesting that other, still-uncharacterized modulators of immune reactivity may determine who reacts and who does not. Larger studies are needed to clarify the true risk and mechanisms [1]. In short: the event is exceptional in clinical practice. And for our particular patient — elderly, volume-sensitive, admitted for heart failure — the most likely transfusion complication would not be anaphylaxis at all, but TACO. The same physiology that brought them into the hospital also raises their risk for fluid overload if transfused. Re-examining the “Requires Washed RBCs” Reflex So where does that leave us? With vigilance, yes — but also with perspective. The patient’s risk for anaphylaxis appears theoretical, not demonstrated. Yet the “washed RBCs” flag carries real-world costs: longer wait times, product scarcity, and potential delays in care. We decided to order an anti-IgA assay to see whether we could safely lift the restriction — a small act of course-correction that might spare the patient unnecessary complexity in future transfusions. Because sometimes the best transfusion practice isn’t about adding more caveats. It ’s about knowing which ones no longer serve the patient. References Rachid R, Bonilla FA. The Journal of Allergy and Clinical Immunology.  2012;129(3):628-34. Williams SJ, Gupta S. Archivum Immunologiae et Therapiae Experimentalis.  2017;65(1):11-19. Salama A et al. Transfusion.  2004;44(4):509-11. Ahrens N et al. Clinical and Experimental Immunology.  2008;151(3):455-8.

Two weeks after a massive transfusion protocol for hemorrhagic shock, one of our patients developed profound thrombocytopenia — counts dropping to single digits despite transfusions. Each additional platelet unit seemed to make things worse, not better. Within days, she developed intra-abdominal bleeding that required surgical exploration and an open abdomen. When the post-transfusion purpura (PTP) panel came back, it revealed an alloantibody against HPA-1b — an uncommon finding, but one that instantly clarified what had happened. What Is Post-Transfusion Purpura? PTP is a rare, delayed transfusion reaction characterized by sudden, severe thrombocytopenia appearing 5–12 days after transfusion. The syndrome occurs most often in individuals who lack the common platelet antigen HPA-1a and have been previously sensitized, typically through pregnancy or prior transfusion. The most common culprit antibody is anti-HPA-1a, but antibodies to other platelet antigens — including HPA-5a, HPA-4a, HPA-3a, and rarely HPA-1b — have been described. Regardless of the specific target, the result is the same: the patient’s immune system destroys both transfused and autologous  platelets, leading to precipitous thrombocytopenia and risk of life-threatening bleeding. A Quick Detour Into HPA Genetics The HPA-1 system is defined by two alleles: HPA-1a, found in roughly 75–95% of most populations. HPA-1b, a minor allele whose frequency ranges from 10–24% in many European and Middle Eastern groups, but is <1% in East Asian populations. This population variation matters. In regions where HPA-1b is rare, like the US, compatible donors for patients with anti-HPA-1b can often be found locally, as most platelet units will be negative for HPA-1b. In the US, sourcing units for patients with anti-HPA-a1 is the reverse — the search for HPA-1b-positive, 1a-negative donors is extraordinarily difficult. Diagnosing PTP PTP should be suspected when: Severe thrombocytopenia develops 5–10 days post-transfusion, There is a paradoxical worsening  of counts after platelet transfusion, and There is no evidence of DIC, HIT, or marrow suppression. Diagnosis is confirmed by detecting platelet-specific alloantibodies in the patient’s serum — most commonly anti-HPA-1a. In this case, however, the antibody was anti-HPA-1b, underscoring that the immunologic mechanism is the same even when the target is rare. Treatment: Why IVIG Works (and Platelets Usually Don’t) The mainstay of treatment is intravenous immunoglobulin (IVIG) — typically 2 g/kg divided over 2–5 days. IVIG acts by saturating Fc receptors, blunting macrophage-mediated platelet destruction, and modulating the immune response. It leads to a platelet recovery in ~85% of reported cases, usually within several days. Plasma exchange can be considered in refractory cases or when IVIG is unavailable, though evidence is limited. What doesn’t  work well is platelet transfusion. Even HPA-matched platelets are often rapidly destroyed in the presence of circulating antibody. Their use is generally reserved for life-threatening bleeding when IVIG has failed or is contraindicated. Reports of transient benefit exist, but consistent success is rare. For long-term management, future RBC transfusions should be washed to remove residual platelets and platelet antigens, and HPA-compatible platelets should be considered to avoid re-exposure and further alloimmunization. Reflection From the Bench This case was a reminder that transfusion medicine sits at the crossroads of immunology and critical care. PTP may be rare, but when it strikes, it can upend a patient’s course and confound even seasoned teams. The irony of PTP is striking: transfused platelets meant to heal instead trigger the destruction of every platelet in circulation. But timely recognition, serologic confirmation, and early IVIG can turn the tide — and save both platelets and lives. When I called hematology with the antibody result today, they immediately pivoted to planning for HPA-matched support before the patient’s next surgery. That’s the value of diagnosis — not just knowing what went wrong, but knowing how to protect the patient the next time around.

When the phone rings mid-transfusion and the words “the patient is hypotensive”  hit your ears, there’s a short list of life-threatening reactions you can’t afford to miss. Four share a similar opening act — fever, hypotension, and sometimes respiratory distress. The fifth looks different but can end the same way if unrecognized. Here’s how to triage the call, fast. 1️⃣ Anaphylaxis Clue:  Sudden hypotension, respiratory distress, flushing, or urticaria — often within minutes  of starting the unit. Ask:  “Did the team give epinephrine?” If they didn’t, that’s step one. Stop the transfusion, keep the line open with saline, and treat per anaphylaxis protocol. Later, you’ll think about IgA deficiency and washed products — but right now, it’s airway, breathing, circulation. 2️⃣ Septic Transfusion Reaction Clue:  Fever and rigors that escalate to shock, often during or shortly after transfusion. Ask:  “Did the team send blood cultures and start antibiotics?” Stop the transfusion immediately and culture both  patient and product. Gram-negative sepsis from a contaminated platelet or red cell unit can mimic anaphylaxis in its speed. 3️⃣ TRALI (Transfusion-Related Acute Lung Injury) Clue:  New or worsening hypoxia and bilateral infiltrates within 6 hours of transfusion, without signs of circulatory overload. Ask:  “Did the oxygen saturation drop or the O₂ requirement go up?” If yes, order a chest X-ray. This is non-cardiogenic pulmonary edema — no diuretics, no fluids, just supportive care and notification to the blood bank. 4️⃣ Acute Hemolytic Transfusion Reaction (AHTR) Clue:  Fever, flank or back pain, hypotension, dark urine, or a sudden rise in bilirubin. Ask:  “Can you send a DAT and haptoglobin?” This one’s about clerical error until proven otherwise. Check patient and unit IDs, call the blood bank, and prepare for aggressive hydration to protect the kidneys. 5️⃣ TACO (Transfusion-Associated Circulatory Overload) Clue:  Hypertension, dyspnea, pulmonary edema, elevated BNP — usually in patients with limited cardiac reserve. Ask:  “Did the team give diuretics, and did the patient respond?” Unlike TRALI, TACO should  improve with diuresis. Prevention is key: slow rates, split units, pre-dose furosemide when indicated. 🩸 Putting It All Together Reaction BP Trend Fever Resp Distress Key Test / Treatment Anaphylaxis ↓ ± + Epinephrine Septic ↓ + ± Cultures + Antibiotics TRALI ↓ ± + CXR → Non-cardiogenic edema AHTR ↓ + ± DAT / Haptoglobin TACO ↑ – + Diuretics → Improves Bottom line: When every minute matters, think pattern-recognition first, paperwork later. Five questions can save a life — and keep you calm when the call comes in.

Every so often a case comes along that refuses to fit into our tidy categories. An adult male presented to the emergency department with profound weakness and striking jaundice. His hemoglobin was 3.7 g/dL, yet he was mentating normally and his lactate was within range — clear evidence of physiologic compensation. The chemistry panel featured a total bilirubin >70 mg/dL, direct fraction ≈ 40 mg/dL, with biliary dilation on CT abdomen/pelvis. On the hematology side, LDH was elevated, haptoglobin undetectable, and the antibody screen revealed a warm autoantibody. Parsing the Pattern At first pass, this looks like hemolysis. Severe anemia, high indirect bilirubin, low haptoglobin, elevated LDH — all the hallmarks are there. But a direct fraction comprising more than half the total complicates the picture: pure hemolysis shouldn’t yield that much conjugated bilirubin. That left two possibilities running in parallel: Primary biliary obstruction (choledocholithiasis, mass, or stricture). Secondary cholestasis from sustained hemolytic load — the liver overwhelmed by the ongoing destruction and turnover of red cells. The patient’s stability pointed toward chronicity. A hemoglobin of 3.7 g/dL with preserved mentation doesn’t happen overnight. This was the physiology of slow, compensated hemolysis finally tipping into hepatic decompensation. Hemolysis Meets Cholestasis In warm autoimmune hemolytic anemia (AIHA), extravascular destruction can be relentless but insidious. Over weeks to months, bilirubin production rises and the liver adapts — upregulating conjugation and secretion. Eventually, canalicular transport capacity becomes the rate-limiting step. The bile becomes supersaturated with bilirubin diglucuronide, predisposing to pigment stone formation and sometimes true biliary obstruction. At that point, the chemistry shifts: direct hyperbilirubinemia emerges, not because the process stopped being hemolytic, but because the liver and biliary tree have been drawn into the downstream pathology. The biliary dilation on CT fits perfectly — not as a primary obstruction, but as a secondary phenomenon of chronic pigment overload. Testing in Context The serologic workup tied it together. Positive DAT with warm autoantibody → ongoing immune hemolysis. Elevated LDH and absent haptoglobin → hemolytic physiology confirmed. Normal lactate and preserved mental status → chronic compensation. The picture that emerged was not an acute biliary event but chronic AIHA complicated by secondary cholestasis, possibly with choledocholithiasis from pigment stones. Biliary Workup MRCP revealed subtle hilar ductal thickening, raising concern for an underlying cholangiocarcinoma. Yet the broader picture didn’t cooperate: tumor markers and an autoimmune cholangiopathy panel were unremarkable. To clarify, an ERCP was performed — several darkly pigmented stones were extracted, and brushings of the bile duct were sent for cytology. No malignant cells were identified. The findings reinforced the idea of secondary  biliary involvement rather than a primary neoplasm. The obstruction appeared mechanical but self-limited, the likely consequence of pigment stone formation in the setting of chronic hemolysis. It was a reminder that when the biliary tree becomes the downstream casualty of hematologic disease, imaging can mimic malignancy and tumor markers can mislead. Takeaway This case reminds me how rarely biology stays within our categorical lines. Direct hyperbilirubinemia does not exclude hemolysis when destruction has been sustained enough to overwhelm excretory capacity. It’s the intersection we rarely see illustrated in textbooks: where hemolysis becomes overwhelming, the liver becomes a bottleneck, and “prehepatic” injury evolves into a mixed cholestatic picture. Sometimes the right answer really is “both.”

The patient was a premature infant, six weeks old and just 2.5 kilograms, already on ECMO for primary cardiopulmonary failure. Sepsis developed secondarily, and the critical-care team requested plasmapheresis for a presumed cytokine storm — a Category III indication  under the current ASFA guidelines. On paper, the rationale made sense. But when I calculated the total blood volume — only 250 mL  (≈ 100 mL/kg for a premature infant) — it was immediately clear this baby was too small for the machine. The Terumo technical support representative confirmed the following minimum specifications  for therapeutic plasma exchange: 30 cm:  Minimum height the system can accept 2 kg:  Minimum weight the system can accept 300 mL:  Minimum total blood volume (manual entry) 10 %:  Minimum hematocrit 0.5 L:  Minimum plasma volume (machine default = 1.0 L) 25 kg:  Minimum weight for automatic TBV calculation Despite meeting the weight requirement, the patient’s total blood volume was below the system’s operational threshold. The procedure was simply not possible. After discussing options with the ICU team, we recommended whole blood exchange  instead — a technically feasible alternative that allowed for cytokine and toxin removal within the constraints of neonatal physiology. This case underscored an important reality: sometimes the limits we face are not physiologic but mechanical. Even when a treatment is conceptually justified, our instruments may not yet be scaled to our smallest and most fragile patients. Until apheresis technology evolves to meet neonatal demands, these moments will remain a quiet reminder that innovation in transfusion medicine isn’t only about what  we can do — it’s also about who  we can safely do it for.

The antibody screen looked straightforward at first glance — an O positive patient with apparent anti-D reactivity in plasma. But the autocontrol was positive, and the eluate was a panagglutinin. Those two results change the entire story. Working the Differential When a D-positive patient’s plasma reacts like anti-D, the immediate differentials are familiar: Partial D variant (missing epitope exposure) Passive anti-D (recent RhIg or IVIG) Autoantibody with apparent Rh specificity Each of these has a characteristic fingerprint. A partial D can produce alloanti-D, but that’s rare in a serologically D-positive person without prior exposure, and the eluate would mirror the plasma pattern, not broaden to panreactivity. Passive anti-D behaves cleanly: plasma reacts like anti-D, eluate is negative or weakly specific, and there’s a clear administration history. What we had was different: a positive autocontrol and a panreactive eluate.That combination eliminates the first two possibilities. The reactivity isn’t alloimmune or passive — it’s autoimmune. The Testing Trail The workup unfolded in layers: Antibody screen — positive with a D-like pattern. Antibody identification panel — strengthened reactivity with D-positive cells, weaker or negative with D-negative cells, suggesting “anti-D.” Autocontrol — positive, establishing that the antibody also reacts with the patient’s own RBCs. Elution study — recovered antibody from patient RBCs showed panreactivity  against all reagent cells. That last result is the pivot point. When an eluate reacts with every cell tested, regardless of Rh type, it means the antibody bound in vivo is not truly D-specific. It’s reacting broadly, most often with epitopes shared across the Rh complex. The apparent anti-D in plasma was the “tip” of the reaction spectrum; concentrating the antibody in the eluate revealed its full range. Understanding the Mechanism This pattern — an apparent anti-D that becomes panreactive on elution — is classic for autoantibodies with relative Rh specificity. Several studies have shown that the Rh complex is a common autoantigenic target in warm AIHA. Barker et al. demonstrated that autoantibodies from AIHA patients could immunoprecipitate Rh-associated polypeptides, confirming that these antibodies truly interact with the Rh complex [1].Later, Iwamoto et al. mapped this reactivity to extracellular peptide loops of Rh antigens (RhD, cE, ce, CE), showing that the apparent specificity stems from how these antigens share structural epitopes [2]. That explains why plasma reactivity can look D-restricted — the antibody’s affinity may be stronger for cells expressing D — but once concentrated, the cross-reactivity becomes universal. Issitt and Pavone and later Henry et al. described this as the “false specificity” of Rh-directed autoantibodies: antibodies that seem to be anti-D, anti-e, or anti-C but are actually targeting shared determinants like Hr, Hro, or Rh34 [3,4]. Clinical Handling In this case, the patient had no evidence of hemolysis and was evaluated in the outpatient setting for routine pretransfusion testing. We were able to rule out any underlying alloantibodies on the work up, and when transfusion was later anticipated, we selected a D-negative unit out of caution. The transfusion was uneventful — confirming that the pattern was serologic, not pathologic. Why It Matters The take-home message isn’t about the D antigen at all — it’s about pattern recognition in serology . When plasma looks specific but the eluate doesn’t, the specificity is probably illusionary. The positive autocontrol and panagglutinin eluate are the signposts pointing away from alloimmunization and toward autoimmunity — specifically, Rh-complex targeting. In documentation, this deserves clarity: “Warm autoantibody with apparent anti-D specificity. Eluate panreactive. No evidence of alloanti-D.” It’s a small notation that prevents a future technologist from re-investigating a mystery that isn’t one — and prevents a clinician from worrying about sensitization that never occurred. References Barker JE et al. J Clin Invest.  1989;84(3):1010–1015. Iwamoto S et al. Blood.  1995;86(1):341–348. Issitt PD, Pavone B. Transfusion.  1978;18(6):702–708. Henry SM et al. Vox Sang.  1987;52(3):193–198.