Staff checking AI suggestions got less accurate: 92.5% down to 81.2%. In a wrong-blood-in-tube study of 510 chemistry result sets, a model working alone reached 92.5% accuracy. When lab staff reviewed its suggestions before deciding, accuracy fell to 81.2%, and sensitivity from 90.6% to 73.7%. The study used a simulated 50% error rate, far above real life, so treat the size of the drop with care. A human in the loop helps only if the loop is designed for it. Where does the human sit in yours? #AIinHealthcare #LaboratoryMedicine #HumanInTheLoop #PatientSafety #AutomationBias

Autoimmunity starts in the womb. Our grandmothers said it long before immunologists did: “What you eat while pregnant and breastfeeding, your child will learn to eat.” They were onto something. 🤰 IN THE WOMB A mother carries a baby who is half foreign, and her immune system chooses not to attack. Pregnancy itself is a lesson in tolerance. Her antibodies cross the placenta to protect the newborn. Her cells prompt the baby to make its own regulatory T cells, the immune system’s peacemakers. The first lesson in tolerance is taught by the mother. 🥛 AFTER BIRTH: THE ENTERO-MAMMARY AXIS Here is the part almost nobody knows. B cells in the mother’s gut lymphoid tissue (MALT) meet everything she eats and every microbe she carries. During breastfeeding, these primed cells travel from her gut to her breast. There they produce secretory IgA into the milk, matched to what her gut has met. Her milk also carries traces of her food proteins, plus calming signals like TGF-β. The baby’s gut receives the message: “This is food. Not a threat.” 🔬 THE EVIDENCE In mice, food proteins passed through breast milk induced tolerance and protected the pups from allergic asthma. In humans, children of non-allergic mothers who ate more peanuts and tree nuts around pregnancy had less nut allergy. The human evidence is still growing, but it points the same way. ⚠️ THE OTHER SIDE The same bridge that carries protection can carry disease. Maternal autoantibodies can cross too: anti-Ro (neonatal lupus), TSH-receptor antibodies (neonatal Graves’), acetylcholine-receptor antibodies (neonatal myasthenia). Most fade within months. Many are predictable, if we test the mother in time. ✅ THE TAKEAWAY • Don’t avoid common foods in pregnancy unless a doctor advises it • Eat a varied diet • Breastfeed where possible • Know the mother’s antibodies Tolerance comes from exposure, not avoidance. A mother’s gut is her baby’s first immunology teacher. Old wisdom. New science. nature.com/articles/nm1718 doi.org/10.1001/jamapediatri… #Autoimmunity #Pregnancy #Breastfeeding #Immunology

Not every hemolyzed sample was hemolyzed in the tube. In a lab thread, a nurse admitted the draw is usually the cause, then added: some patients are on devices that break down their own red cells. They are right. A 2025 review in the Journal of Applied Laboratory Medicine lists genetic conditions, hemoglobin disorders and mechanical circulatory support as causes of hemolysis inside the patient, and notes that routine lab analysis cannot tell it apart from tube hemolysis. If clean redraws keep coming back hemolyzed, the finding may be the patient, not the sample. #BeyondTheSample #Hemolysis #LaboratoryMedicine

Before you call the lab, spend a shift in it. In a nursing thread with more than 3,500 upvotes, a nurse proposed letting teams that depend on each other shadow one another. The reply: suggested at every hospital, everyone thinks it's a great idea, "except leadership." A new graduate nurse in a lab thread said the same: "I would love shadowing lab for a day." Training across that line works. After a phlebotomy training program in one emergency department, troponin hemolysis fell from 7.2% to 3.4% in twelve weeks. #InterprofessionalEducation #LabAndNursing #PatientSafety

When the lab cancels a test, does anyone on the ward find out? One nurse described it plainly: at their hospital the lab does not call, cancelling on the lab's side triggers no notification, so they keep refreshing, waiting for a result that is never coming. A cancellation is an action in the lab system like any other. A rule can send an alert with it, the same way a rule can hold a hemolyzed potassium or cancel a duplicate. Where a result stops, someone should be told. #LabCommunication #PatientSafety #CloseTheLoop

The clinician saw one lesion. The microscope saw another. In a study of 910 oral lesions, the clinical impression and the histologic diagnosis disagreed about 1 in 4 times (25.1%). Doctors answering a worried patient online said the same thing: some lesions look obvious through the scope, and the biopsy still has to confirm. When the two disagree, where does that conversation happen in your hospital? #OralPathology #DiagnosticAccuracy #ClinicalCollaboration

If the specimen is not on the validated list, the answer is no. Then pick up the phone. A night-shift tech was recently handed breast milk for a cell count. Colleagues' advice: call the doctor, explain it is not an acceptable fluid, and cancel it. Under US rules, a lab that changes a cleared test, such as running it on a new sample type, must first establish accuracy, precision, analytical sensitivity and specificity, reportable range and reference intervals. A cancelled test is frustrating. A confident wrong result gets acted on. #LabQuality #SpecimenValidation #PatientSafetyFirst

A national lab rule re-run on 9 million results fit most labs. For three labs, it barely fit. England built one acute kidney injury warning algorithm into every lab system. Researchers re-ran it centrally on 9,096,667 creatinine results from 29 labs. Agreement with each lab's own alerts was 0.97 to 0.98 in most, and as low as 0.17 to 0.23 in three. Local edits and suppression had changed the output. Pooled lab data is only as good as each lab's build, so before you compare labs, ask where the rule was changed. #LabQuality #ClinicalInformatics #AcuteKidneyInjury

GANGLION CELL OR SCHWANN CELL? ONE MISTAKE CAN CHANGE A CHILD’S DIAGNOSIS! In #HirschsprungDisease, not every stained cell is a ganglion cell! Ganglion cells: Larger neuronal bodies, vesicular nuclei, prominent nucleoli and abundant cytoplasm. Schwannian cells: Small, elongated nuclei, scant cytoplasm, associated with nerve bundles. The immunostain trap: • Calretinin: Highlights many ganglion cells and intrinsic mucosal nerve fibers. • SOX10/S100: Highlight Schwannian/glial elements—not proof of ganglion cells. • HuC/D: Helps identify neuronal cell bodies. The golden rule: A positive stain is NOT a diagnosis. Morphology + location + immunophenotype matter! #Pathology #PediatricPathology

Everyone asks if the AI was accurate. Ask what it could not see. In a structured test of a consumer AI health tool on 960 responses, its errors peaked at the extremes: 35% in nonurgent cases and 48% in emergencies. When family members in the scenario played down the symptoms, its advice shifted toward less urgent care in edge cases, with an odds ratio of 11.7. A tool that only hears the story inherits the story's blind spots. #AIHealthcare #ClinicalAI #PatientSafety

"It wasn't hemolyzed when I sent it" may well be true. In a lab thread about a nurse who redrew three times, all hemolyzed, that line was the top comment. Hemolysis can start after the needle: in the pneumatic tube, in a long wait before spinning, in rough handling. Where to look: • collection time against receipt time • the transport route • how long the tube waited before the centrifuge The timestamps usually know. #LabQuality #PreanalyticalErrors #PatientSafety

ESOPHAGEAL CANCER AT 24. DID THE STORY BEGIN AT 10? Recently, we encountered a 24-year-old with esophageal adenocarcinoma, extensive high-grade dysplasia and Barrett’s-type changes. An extraordinary diagnosis at this age! But one question keeps bothering me. What if chronic reflux had started when he was only 10? We don’t know his childhood history. But imagine a severely obese child suffering from persistent reflux for years. Who follows that child? Do pediatric gastroenterology services have sufficient capacity? Is upper GI endoscopy accessible and covered when clinically indicated? Are we recognizing warning symptoms early enough? Not every child with reflux needs endoscopy. Most will never develop cancer. But persistent reflux should never be ignored simply because the patient is young. Cancer diagnosed at 24 may raise questions about healthcare opportunities missed years earlier. #GERD #BarrettsEsophagus #EsophagealCancer #ChildhoodObesity #PublicHealth

If an AI writes the preliminary report, does a pathologist still have to look at the slide? A lab worker asked this online after a pathologist at another site signed out biopsies without seeing the slides. One reply: "An extremely relevant question in the age of AI." Others pointed to inspections: a qualified pathologist signs out the case, with peer review behind it. My answer: a signature means someone looked. An AI preliminary report can change the order of the work, not who is responsible for it. #AIInPathology #HumanOversight #PatientSafetyFirst

61.2% of request forms had no contact number. So who gets the critical value? In a study of 2,550 forms, 151 critical results came with incomplete ward information, and 19.9% of those were never phoned. Picture the night tech holding a potassium of 6.4, a form that says only "Ward 4," and no name to ask for. The lab cannot call faster than the form allows, so every order needs a reachable name and number. #CriticalResults #PatientSafety #LabMedicine

ONE RARE OVARIAN TUMOR. TWO DIFFERENT DIAGNOSES. WHICH ONE IS RIGHT? A 34-year-old woman with an 8 cm ovarian mass. Initially diagnosed as Sertoli–Leydig cell tumor. International pathology review later favored juvenile granulosa cell tumor (JGCT). But this was NO textbook JGCT! Under the microscope, TWO striking patterns: Solid, densely cellular lobules with irregular follicles, variable-sized cystic spaces and eosinophilic secretions. Loose spindle-cell, fibrovascular and thecomatous-appearing areas, with focal tubular-like structures mimicking Sertoli–Leydig differentiation. Even the immunostains were confusing, showing unexpected heterogeneous staining patterns! Yet no convincing separate Sertoli–Leydig component was identified in our reviewed sections. THE PATHOLOGY PEARL: Two morphological patterns do NOT necessarily mean two different tumors. Juvenile granulosa cell tumors can have many faces. And distinguishing these rare tumors matters—because diagnosis influences staging, treatment decisions, fertility preservation and long-term follow-up. Sometimes morphology must solve the mystery that immunohistochemistry creates. #Pathology #OvarianTumor #GranulosaCellTumor #Histopathology

The most trusted call a nurse ever got from the lab: "I dropped your sample." In a nursing thread with more than 3,500 upvotes, that was the top comment. A lab tech called to say the tube was dropped and spilled, and the nurse respected the truth far more than any excuse. In our hospital, tubes travel through the pneumatic system in dedicated lab capsules, and a misrouted or late capsule can still ruin a specimen. Trust grows when the lab says where a failure happened, even when the answer is the lab. #TransparencyBuildsTrust #LabLife #PatientSafety