How VTE Prophylaxis Decisions Are Risk Stratified in the Hospital

Ordering a blood thinner for a hospital patient is rarely a simple yes or no. It is a balance between two harms that pull in opposite directions: the clot a patient might form if left unprotected, and the bleed that same patient might suffer if given an anticoagulant. Risk stratification is the discipline of estimating each harm separately, then acting on whichever one is larger and more dangerous. Validated scores put numbers on both sides, and the American Society of Hematology (ASH) 2018 guideline converted that comparison into graded recommendations for medical inpatients.

Key points#

• VTE prophylaxis is a trade-off between two competing risks, clot and bleed, that are scored separately rather than as one number.
• The Padua and IMPROVE VTE scores estimate clot risk; the IMPROVE bleeding score estimates bleeding risk.
• A practical sequence checks bleeding risk first, then offers a drug only when clot risk is high and bleeding risk is acceptable.
• The ASH 2018 guideline made most of these calls conditional recommendations, meaning the balance is close and depends on the individual.

Why not just treat everyone#

It is tempting to think that if clots are dangerous, every admitted patient should get prophylaxis. The arithmetic argues otherwise. Most medical inpatients will never form a clot, and the drugs used to prevent one, chiefly low-molecular-weight and unfractionated heparins, carry a real bleeding cost of their own. Treating an entire ward would subject a large low-risk group to bleeding harm in order to protect a small high-risk group. Because hospital-associated venous thromboembolism arises heavily in non-surgical patients, the ASH panel wrote a whole guideline on medical inpatients. The purpose of stratification is to identify the minority who benefit before anyone is treated.

Scoring the clot side#

The Padua Prediction Score#

The Padua Prediction Score comes from a 2010 prospective cohort by Barbar and colleagues in the Journal of Thrombosis and Haemostasis. It totals points across eleven items. Four factors weigh heaviest at three points each: active cancer, previous venous thromboembolism, reduced mobility, and known thrombophilia. Recent trauma or surgery adds two. Seven lower-weight factors add one point each: age of seventy or older, heart or respiratory failure, recent stroke or myocardial infarction, acute infection or rheumatologic illness, obesity, and current hormonal treatment. A total of four or more classes the patient as high risk.

What makes the cutoff worth using is the gap it reveals. In the derivation cohort, high-risk patients left without prophylaxis went on to develop symptomatic clots roughly eleven percent of the time. High-risk patients who were treated sat near two percent, and low-risk patients near 0.3 percent. One number separated a group with a meaningful clot rate from a group in whom a drug would mostly add danger without payoff.

The IMPROVE VTE score#

Padua is not the only clot-risk tool. The IMPROVE program, the International Medical Prevention Registry on Venous Thromboembolism, produced its own VTE score built on predictors including prior clot, known thrombophilia, lower-limb paralysis, active cancer, prolonged immobilization, intensive care admission, and older age. Both tools answer the same question, how likely is this patient to clot, and either can anchor the clot side of the decision.

Scoring the bleeding side#

A clot score alone is only half a decision. The companion IMPROVE bleeding risk score estimates the opposite danger. It draws on factors that predict hemorrhage rather than thrombosis: active gastroduodenal ulcer, recent bleeding, a low platelet count, advanced age, liver failure, severe kidney impairment, an intensive care stay, a central venous catheter, and active cancer, which notably raises both risks at once. When a patient's bleeding score reaches its threshold, an anticoagulant is more likely to cause harm than to prevent it, and that finding can override an otherwise elevated clot score.

Putting the two scores together#

The scores earn their keep when read as a pair, not in isolation. A 2024 analysis by Djulbegovic and colleagues in Blood Advances showed how the IMPROVE VTE and bleeding models can be arranged into a short sequential decision tree. The sequence checks bleeding risk first. If it is high, pharmacologic prophylaxis is withheld regardless of the clot number. Only when bleeding risk is acceptable does the tree look at clot risk and offer a drug to those whose clot risk is genuinely elevated. Ordering the steps that way encodes an old clinical instinct: rule out the harm you might cause before chasing the benefit you hope for.

What the ASH 2018 panel recommended#

The ASH 2018 guideline, developed with the McMaster GRADE Centre, did not reduce this to one universal rule. It issued graded recommendations that map onto the two-sided assessment.

• For acutely and critically ill medical inpatients, it suggested pharmacologic prophylaxis over none, on the reasoning that clot risk usually outweighs bleeding risk in this group.
• When a drug is chosen, it favored low-molecular-weight heparin over unfractionated heparin, citing once-daily dosing and fewer complications.
• It preferred inpatient low-molecular-weight heparin over continuing a direct oral anticoagulant after discharge, declining to endorse routine extended prophylaxis.
• For patients at high bleeding risk who still need protection, it preferred mechanical prophylaxis, such as pneumatic compression, over drugs.
• For patients at high clot risk but acceptable bleeding risk, it preferred the drug over mechanical measures.
• It judged combined mechanical-plus-drug prophylaxis unnecessary for most medical inpatients, since the extra burden did not buy clear extra benefit.

Most of these are conditional under GRADE. That wording is a deliberate signal: the balance is close, and a reasonable clinician and patient might weigh it differently.

The limits of any score#

A score is a starting estimate, not a verdict. The ASH panel framed its advice around clot and bleeding risk without declaring any single tool mandatory, and a model derived in one population can behave differently in another. A score also describes an average patient, while the person in the bed may sit at the edge of the distribution the model was built on. The real value of Padua and IMPROVE is procedural: they force both halves of the question, the clot and the bleed, to be named and weighed against each other, so that neither side goes unexamined before you decide.