Blood Transfusion Thresholds
Blood is a transplant. It has a donor, an immune consequence and a dose, and the most common error on the wards is giving two units to a number instead of one unit to a patient.
Product availability, crossmatch turnaround, irradiation and leukoreduction policy, massive transfusion protocol contents and consent forms are all local. Everything below is the general shape of the practice. The operative documents are your institution's protocol and a current reference.
The restrictive strategy and why it won
For most hospitalized adults the transfusion threshold is a hemoglobin of 7 g/dL. Not 8 because the patient looks pale. Not 9 because they are going to the operating room tomorrow. Not 10 because that is the number you learned as a student.
This was not obvious. For decades the reflex was to keep everyone near 10, on the reasonable sounding theory that more oxygen carrying capacity is better. Then people ran the trials. TRICC, in critically ill adults, found a restrictive strategy at least as good as a liberal one, with a signal toward better outcomes in the younger and less sick. TRISS asked the same question in septic shock and found no mortality difference. FOCUS looked at hip fracture patients with cardiovascular disease and found no benefit from the liberal threshold. Across surgery, gastrointestinal bleeding, cardiac surgery and general medicine the pattern repeated: more blood does not make patients live longer, and in several settings it makes them do worse.
The reasons are not mysterious. Stored red cells are depleted of 2,3-DPG, their membranes are stiffer, and they do not offload oxygen the way circulating cells do. Every unit is also an immune exposure and a volume load. A liberal strategy buys a prettier lab value and pays for it with circulatory overload, febrile reactions, antibodies that make the next crossmatch harder, and cost. So restrictive is the default, and the burden of proof sits on whoever wants to transfuse higher. That person needs a reason they can say out loud and write down.
A hemoglobin of 6.8 in a comfortable, ambulating patient with iron deficiency and no bleeding is an argument for iron, not for a unit of blood. A hemoglobin of 8.2 with chest pain, ST depression and ongoing melena is an argument for blood. The number opens the conversation. It does not end it.
The threshold table by population
| Population | Hemoglobin | What actually drives the decision |
|---|---|---|
| General medical or surgical inpatient, hemodynamically stable | 7 | The default. Covers most of your list, including sepsis and the ICU. |
| Acute coronary syndrome, active ischemia | 8 to 10 | Genuinely unsettled. Most land near 8, higher with ongoing ischemia. |
| Cardiac surgery, intraoperative and postoperative | 7.5 | Restrictive has held up here, including on bypass. |
| Orthopedic surgery with cardiovascular disease | 8 | Or symptoms. The hip fracture population is where the 8 comes from. |
| Upper gastrointestinal bleeding | 7 | Better, not merely equivalent. In varices it is actively protective. |
| Oncology, chemotherapy induced or chronic anemia | 7 to 8 | Symptom driven. Find the treatable cause first. |
| Sickle cell disease | Situational | Not a threshold at all. Driven by the complication. Above 10 is a hazard. |
Acute coronary syndrome: the honest version
This is the one population where the restrictive answer is not settled, and you should know that rather than quoting 7 with false confidence. The physiologic argument for a higher threshold is real: an ischemic myocardium is fed by a stenosed vessel, extraction is already near maximal, and anemia removes the last reserve. The counterargument is also real, since transfusion is not benign and the older cardiac trials did not show the benefit the physiology predicted.
The MINT trial addressed this directly in myocardial infarction with anemia, comparing restrictive against a liberal threshold near 10. The result was mixed: the primary outcome did not reach statistical significance, but the point estimates favored the liberal arm and the cardiac specific outcomes leaned the same way. It does not prove liberal transfusion works, and it does not let you comfortably run an active infarct at 7.2. A threshold around 8 is defensible, and going higher with ongoing ischemic pain or dynamic electrocardiographic changes is defensible. Say in the note which way you went and why.
Upper gastrointestinal bleeding, and the variceal trap
Here a threshold of 7 is not merely non-inferior, it is better, and the effect is largest in cirrhosis. The mechanism matters because it changes what you do at three in the morning. Volume expansion, blood or crystalloid, raises splanchnic blood volume, which raises portal pressure, which raises the pressure inside the varix that just stopped bleeding. Resuscitate a variceal bleeder to 10 and you can blow the clot off. The classic sequence is a patient who stabilized, got four units toward a number that pleased somebody, and rebled worse.
Target 7 to 8. Do not chase a normal hemoglobin. Give octreotide and antibiotics, get the endoscopist, and resist the pressure to keep hanging units because the number looks bad. If the patient is exsanguinating in front of you that is a different situation and you transfuse to keep them alive. Ongoing hemorrhage overrides every threshold on this page.
Oncology and sickle cell
Chronic anemia is compensated anemia. The patient on chemotherapy at 7.4 who walks to the bathroom without symptoms needs a cause identified, and often iron or an erythropoiesis stimulating agent, rather than a transfusion that lasts three weeks and produces antibodies.
Sickle cell is where a threshold is the wrong mental model entirely. These patients live at a baseline that would alarm you in anyone else, and transfusing a vaso-occlusive crisis to improve the number does nothing for the pain while raising viscosity. Transfusion here is for specific indications: acute chest syndrome, stroke and stroke prevention, acute symptomatic drops such as splenic sequestration or aplastic crisis, and preoperative optimization for certain procedures. Simple transfusion for a modest correction, exchange transfusion when the goal is to lower the hemoglobin S percentage without raising viscosity. A post-transfusion hemoglobin above roughly 10 is a hyperviscosity hazard. Use extended phenotype matched units, because these patients are transfused for life and alloimmunization is what eventually makes them uncrossmatchable. Involve hematology early.
One unit at a time, then recheck
One unit of packed red cells raises the hemoglobin by roughly 1 g/dL, or the hematocrit by about 3 points, in a 70 kg adult who is not actively bleeding. That is the whole calculation.
Which means the reflex two unit order is a habit rather than a decision. The patient at 6.8 whom you want above 7 needs one unit. You gave two because two is the order set default and it saves a phone call later. That second unit is a second immune exposure, a second volume load and a second chance at a reaction, given for no stated reason.
Order one unit. Recheck the hemoglobin. In a patient who is not bleeding, equilibration is quick and a level drawn after the unit finishes is interpretable. Then decide again. The exceptions are few: active hemorrhage, a massive transfusion protocol, and a patient so profoundly anemic that one unit will not get them close. Outside those, "two units" written without a reason is one of the easiest things to criticize in a chart.
The order and the plan
"Transfuse 1 unit PRBC for hemoglobin 6.6 with exertional dyspnea. Recheck CBC after transfusion. Will reassess symptoms and repeat hemoglobin before considering a second unit. No active bleeding. Furosemide 20 mg IV with the unit given heart failure with reduced ejection fraction."
Two practical points travel with the order. Standard infusion is one and a half to two hours per unit, four hours maximum for infection control, faster only if the patient is bleeding. And in heart failure or advanced kidney disease, slow the unit and consider a diuretic with it, because circulatory overload is the most common serious transfusion reaction and it is almost entirely preventable.
The other products
Platelets
Platelet thresholds are about the bleeding risk of the moment, not a target count.
Below 10,000 in a stable patient with hypoproliferative thrombocytopenia, typically after chemotherapy or in marrow failure. Many centers use 20,000 with fever, sepsis or a fast falling count.
50,000. Central line, chest tube, paracentesis, endoscopy with intervention, most surgery.
Lumbar puncture and epidural placement, 50,000 as a common floor, with many anesthesia services wanting more and a stable or rising trend first. Ask them, do not assume.
100,000. Bleeding into a closed box justifies the higher bar.
One apheresis unit, the standard adult dose, raises the count by roughly 30,000 to 50,000. If it does not, think about consumption, splenic sequestration, fever, drugs or alloimmunization. Do not transfuse platelets in heparin induced thrombocytopenia or thrombotic thrombocytopenic purpura unless there is life threatening bleeding, because you are feeding a thrombotic process. The HIT case is on the heparin drip page.
Fresh frozen plasma, and the ritual you should stop performing
FFP is genuinely indicated in a short list of situations: active bleeding with a documented multiple factor coagulopathy, massive transfusion as part of the ratio, warfarin associated major bleeding when PCC is unavailable, plasma exchange for thrombotic thrombocytopenic purpura, and rare factor deficiencies with no concentrate available.
It is not indicated for the thing it is most often ordered for: normalizing a mildly elevated INR in a non-bleeding patient before a bedside procedure. That is a ritual. The evidence does not show it reduces bleeding, and the reasoning is wrong in the population where it is used most.
The INR measures procoagulant factors made by the liver. Cirrhosis lowers the anticoagulant factors, protein C, protein S and antithrombin, in parallel. The result is rebalanced hemostasis at a new and fragile set point, and an INR of 1.8 in a cirrhotic does not mean what an INR of 1.8 on warfarin means. Those patients also form clots.
Two facts end the argument. FFP itself has an INR of roughly 1.6, so you cannot get below that with plasma no matter how much you give. And the dose that does anything is 10 to 15 mL/kg, three or four units and close to a liter of volume in an average adult. Circulatory overload from prophylactic plasma is a real and stupid way to hurt a patient.
Cryoprecipitate
Cryoprecipitate is the concentrated source of fibrinogen, and also carries factor VIII, factor XIII, von Willebrand factor and fibronectin. Give it for a fibrinogen below about 150 mg/dL in a bleeding patient, with obstetric hemorrhage generally managed to a higher target closer to 200. A typical adult dose is a pooled preparation of about ten units, raising fibrinogen on the order of 50 to 100 mg/dL. Check a fibrinogen in any bleeding patient who has received several units of red cells. Dilutional hypofibrinogenemia is common, easy to miss and easy to fix.
Prothrombin complex concentrate
Four factor PCC is the agent for warfarin associated major bleeding, or for an urgent procedure in an anticoagulated patient. It is dosed by weight and INR, comes in a small volume, and works in minutes rather than the hours plasma takes. Give it with intravenous vitamin K, because PCC corrects the INR faster than it corrects the underlying problem and the number rebounds as the factors are consumed. Dosing and the reversal agents for the direct oral anticoagulants are on the warfarin and reversal page.
Massive transfusion protocol
Activate early. The common failure is not over-activation, it is the resident who keeps ordering two units at a time while the patient exsanguinates and the blood bank has no idea anything unusual is happening. Triggers vary, but practically: an anticipated need for large volumes fast, hemorrhagic shock not responding to initial units, a penetrating mechanism with instability, or bleeding you cannot control at the bedside. Say the words out loud and assign one person as the sole point of contact for the blood bank.
The ratio concept. A patient bleeding whole blood who receives only red cells becomes dilutionally coagulopathic. The answer is plasma and platelets alongside the red cells in a balanced ratio approximating 1:1:1. Trial data comparing 1:1:1 against a plasma sparing 1:1:2 found no mortality difference at 24 hours or 30 days, but more patients achieved hemostasis and fewer died of exsanguination early with the balanced ratio. The protocol exists so products arrive together in coolers and nobody computes anything mid-resuscitation.
Blood products are anticoagulated with citrate, and citrate chelates ionized calcium. Transfuse fast and the ionized calcium falls, sometimes precipitously, because the liver cannot clear the citrate load. Hypocalcemia causes hypotension, reduced contractility and worse coagulopathy, so the patient gets more blood, which makes the calcium worse. You can watch that loop close in front of you.
Check an ionized calcium early and repeatedly, and replace empirically rather than waiting for a result. Calcium chloride carries about three times the elemental calcium of the same volume of calcium gluconate but is caustic and belongs in a central line. Gluconate is the peripheral choice. Add hypocalcemia to the list you are fighting alongside hypothermia and acidosis, and warm the products and the patient.
Tranexamic acid. In traumatic hemorrhage, TXA given early reduces death from bleeding, and the benefit is time dependent: earliest is best, and late administration appears harmful rather than merely useless. The usual regimen is 1 g intravenously over 10 minutes then 1 g over 8 hours. It has an established role in postpartum hemorrhage as well. Do not push the load fast, because that causes hypotension.
Whole blood. Many trauma centers and prehospital services have moved back to low titer group O whole blood as the first product in hemorrhagic shock. The logic is simple: the patient is losing whole blood, so replacing it with whole blood delivers a balanced ratio in one bag with less anticoagulant and additive solution. Whether your center stocks it, and who may receive it, is a local question worth answering before the trauma page goes out.
Transfusion reactions
For nearly every reaction below, the first two actions are identical and immediate: stop the infusion and keep intravenous access open with normal saline through new tubing. Do not pull the line, because you may need it in ninety seconds. Then assess the patient, then check the identification on the unit against the patient's wristband, then call the blood bank and send back the unit and the required samples. Diagnosis comes after those steps, not before.
| Reaction | Timing | Presentation | Immediate action |
|---|---|---|---|
| Febrile non-hemolytic | During, or within about 4 hours | Temperature rise of a degree or more, chills, rigors, no hypotension, no hemolysis. The most common reaction. | Stop. Saline. A diagnosis of exclusion, so rule out hemolysis and contamination first. Acetaminophen. Do not restart that unit. |
| Acute hemolytic | Minutes, often within the first 50 mL | Fever, chills, flank or back pain, pain at the infusion site, hypotension, dark urine, oozing from puncture sites. Usually ABO incompatibility from a clerical error. | Stop immediately. Saline wide open. Clerical check at the bedside. Send unit, tubing and post-transfusion samples for direct antiglobulin test and repeat crossmatch. Support pressure, maintain urine output, watch for DIC and hyperkalemia. |
| TRALI | Within 6 hours | Acute hypoxemia with bilateral infiltrates and no evidence of circulatory overload. Often fever and hypotension. Filling pressures not elevated. | Stop. Oxygen and lung protective ventilation as needed. Supportive care only. Diuretics do not help and can hurt. Report it so the donor is investigated and deferred. |
| TACO | During, or within 6 to 12 hours | Dyspnea, orthopnea, hypertension, elevated jugular venous pressure, rales, rising natriuretic peptide. Favors the elderly, small, and cardiac or renal patient. | Stop or slow. Sit the patient up. Oxygen. Intravenous furosemide. It responds to diuresis, which is the practical way to separate it from TRALI at the bedside. |
| Allergic, urticarial | Minutes to hours | Hives and itching without systemic features. | Pause. Antihistamine. If mild and it resolves promptly, this is the one reaction where the unit may be resumed under supervision per local policy. |
| Anaphylactic | Within minutes | Hypotension, bronchospasm, stridor, angioedema, often without fever. Consider IgA deficiency with anti-IgA antibodies. | Stop permanently. Intramuscular epinephrine. Airway, fluids, full anaphylaxis management. Future products may need to be washed or IgA deficient. |
| Delayed hemolytic | 3 to 14 days, sometimes longer | Unexplained fall in hemoglobin after an uneventful transfusion, mild jaundice, rising bilirubin and LDH, positive direct antiglobulin test, new alloantibody. | Nothing to stop. Notify the blood bank so the antibody is recorded, give the patient a card, and get it into the discharge summary. This is what makes future crossmatching hard. |
Two more worth naming. A septic transfusion reaction, from a contaminated unit, brings high fever, rigors and shock during or shortly after the transfusion and is more associated with platelets, which are stored at room temperature: stop, culture the patient and the bag, start broad spectrum antibiotics. And a hypotensive reaction, isolated hypotension without other features, occurs more often on ACE inhibitors.
Consent and documentation
A transfusion is a procedure with a consent conversation, and the note should read like a procedure note rather than a checkbox. What makes it defensible is not length. It is that a reader two years from now can see why you did it, what you told the patient, and what happened after.
"Hemoglobin 6.4, down from 8.9 yesterday, with exertional dyspnea and tachycardia to 112." Not "anemia."
One sentence, if you departed from 7. "Threshold of 8 given active NSTEMI with ongoing chest discomfort."
And the reason, if it is more than one.
Who you spoke with, that risks including fever, allergic reaction, circulatory overload, hemolysis and the small risk of transmitted infection were discussed along with alternatives, that questions were answered, and that the patient or surrogate agreed. If the patient lacks capacity, say how you determined that and who consented instead.
Post-transfusion hemoglobin, whether symptoms improved, whether there was a reaction. A transfusion with no documented response is an intervention nobody evaluated.
A transfusion note that holds up
"Hemoglobin 6.4 this morning, down from 8.9, with new exertional dyspnea and resting heart rate 110. Stool guaiac positive, no overt bleeding. Discussed transfusion at the bedside including risks of febrile and allergic reaction, circulatory overload, hemolytic reaction and the small risk of transmitted infection, as well as the alternative of intravenous iron with slower correction. Questions answered. Patient consents. Transfusing 1 unit PRBC over 2 hours with furosemide 20 mg IV given ejection fraction 35 percent. Will recheck CBC post transfusion and reassess before further units. GI consulted for endoscopy."
Later: "Post transfusion hemoglobin 7.5. Dyspnea resolved, heart rate 88. No transfusion reaction. Holding further units."
For how the wording of a note reads years later, and for the specific verbs that create problems, see medical legal charting.
Jehovah's Witness patients
You will have this conversation, and the way to do it badly is to treat it as an obstacle. A competent adult may decline any treatment, including one you believe will save their life. Your job is to make sure the decision is informed, document it precisely, and then give the best care available inside the boundary they have drawn.
Have the conversation early, before the crisis, and have at least part of it with the patient alone. Not because the family is doing anything wrong, but because a decision this consequential should be verified without an audience. Assess capacity as you would for any refusal of major therapy.
Then get specific, because "no blood" is not one decision. Most patients decline whole blood, red cells, platelets, plasma and white cells. Many treat derivatives and fractions, including albumin, immunoglobulins, clotting factor concentrates and topical hemostatic agents, as a matter of personal conscience and accept some and not others. Many accept procedures where their blood stays in a continuous circuit with their body, which makes cell salvage, acute normovolemic hemodilution and cardiopulmonary bypass often acceptable. Ask about each category, write down each answer, and scan the advance directive card most of these patients carry.
What you have to offer is more than people assume. Intravenous iron rather than oral, dosed properly and repeated. Erythropoiesis stimulating agents with enough iron behind them. Tranexamic acid. Aggressive minimization of phlebotomy, including pediatric tubes and cancelling the daily labs nobody reads. Early surgical or endoscopic control of the bleeding source. Cell salvage where surgery is involved. And tolerating a hemoglobin far below where you are comfortable while protecting oxygen delivery: keep them warm and still, treat pain and fever, supplement oxygen, avoid the tachycardia that raises myocardial demand. Involve hematology and your bloodless medicine service on day one rather than day four.
The word matters more than it should. "Refused" reads as a confrontation and colors every subsequent reader against the patient. "Declined" states the identical clinical fact and describes a decision rather than a fight. It applies to every refusal in the chart, and it matters most here, where the decision rests on a deeply held belief and the note may be read by people looking for someone to blame.
Documenting the decision
"Discussed anemia management with the patient at the bedside, family briefly excused at my request. Hemoglobin 6.1 with ongoing gastrointestinal losses. Explained the indication for red cell transfusion and the risks of remaining anemic, including cardiac ischemia, worsening heart failure and death. The patient declined transfusion of whole blood, red cells, platelets and plasma on religious grounds. She will accept albumin, clotting factor concentrates and intraoperative cell salvage provided the circuit remains continuous. She was consistent on repeat discussion, articulates the consequences accurately, and has decision making capacity. Advance directive card scanned into the chart. Plan: intravenous iron sucrose, erythropoietin, tranexamic acid, phlebotomy minimized to twice weekly with pediatric tubes, GI consulted for urgent endoscopy, hematology for bloodless management. Family aware and supportive of her decision."
Two situations need help beyond you. A patient without capacity whose prior wishes are documented is a different legal analysis than one whose wishes are reported secondhand, and a minor is an entirely different framework in most jurisdictions. Call ethics and institutional counsel before the situation is urgent, not during it.
Related: warfarin and reversal for the anticoagulated bleeder, and IV fluids for what is running alongside the blood.
This is not medical advice. It is a teaching outline for clinicians and clinicians in training. The thresholds, doses and products below are illustrative and simplified. Follow your institution's transfusion policy, verify everything against a current reference, and use your own judgment.