THRIVE Score for Stroke Outcome
The THRIVE score can help physicians predict several key outcomes in patients suffering an ischemic stroke.
The Total Health Risk In Vascular Events (THRIVE) score uses NIHSS score, age, and chronic disease to predict long-term neurologic outcomes in stroke patients.
- Scored on a 0-9 point scale, lower is better.
- A score of 0 predicts a 79-88% chance of a good neurological outcome and 0-2% predicted mortality at 90 days.
- A score of 9 predicts a 7-16% chance of a good neurological outcome and 38-58% mortality at 90 days.
- Note: A newer outcome measure is now performed using the THRIVE-c calculation, which uses continuous age and NIHSS to more accurately determine outcome probability.
- Risk of hemorrhagic conversion increases proportionally for each additional point in the THRIVE score.
- The THRIVE score performs well when applied to stroke patients who received IV tPA as well as patients who did not receive thrombolysis or endovascular intervention.
Points to keep in mind:
- The THRIVE score has been validated only retrospectively in various stroke databases and has not been applied prospectively.
There are nearly 800,000 cases of acute stroke in the United States every year, with 130,000 associated deaths (4th leading cause of death in Americans).
The THRIVE score can help physicians predict functional outcome, death after stroke, and the risk of brain hemorrhage after IV tPA administration in patients who suffer an ischemic stroke.
Result:
Please fill out required fields.

From the Creator
We built the THRIVE score to put real data behind the notion that chronic medical comorbidities can significantly influence ischemic stroke outcomes. It was known that NIHSS and age were major predictors of stroke outcome, so we combined these elements with comorbidities in the score-building data set (the MERCI and Multi-MERCI trials). Three comorbidities were independent predictors in the models : hypertension, diabetes, and atrial fibrillation. Each comorbidity accounted for approximately the same risk of worsened outcome. We went on to validate the THRIVE score in the Merci Registry, the NINDS IV tPA trial, VISTA, the TREVO-2 trial, the SWIFT and STAR trials, and the SITS-MOST study. We found that the THRIVE score works well across the three major acute treatment contexts (IV tPA, endovascular stroke treatment, and no acute treatment) and that THRIVE predicts functional outcome, death after stroke, and the risk of brain hemorrhage after IV tPA administration.
Later work was done to allow THRIVE to better estimate individual patient outcomes. Most clinical scoring systems, like the original THRIVE score, take in some inputs that are dichotomous or binary (e.g., hypertension = YES / NO) and others that are continuous like age. For these continuous variables, traditional scores typically break down these continuous variables using thresholds to make score generation possible (for example, the original THRIVE score assigns 0 points for age ≤ 59, 1 point for age 60-79, and 2 points for age ≥ 80). The continuous THRIVE-c equation uses the continuous inputs of age and NIHSS and combines them with the dichotomous inputs for hypertension, diabetes mellitus, and atrial fibrillation in a mathematical equation that directly estimates outcome probability for individual patients.
The most recent work on THRIVE used this same framework of individual-patient predictive equations to estimate the chances of a good outcome in patients with large vessel occlusion (LVO), stratified by whether they undergo endovascular treatment (EVT). To do this, a series of THRIVE-EVT equations were generated and validated using data from 7 RCTs of EVT (MR CLEAN, SWIFT-PRIME, ESCAPE, EXTEND IA, MR RESCUE, THRILL, and THRACE).
The current version of the THRIVE calculations on MDCalc let the user to input age, NIHSS, hypertension, diabetes mellitus, atrial fibrillation, presence of an LVO (if known), and ASPECTS (if available). Based on the presence or absence of an LVO and the availability of ASPECTS, the correct equation is used, and outcome estimation is better tailored for the individual patient.
About the Creator
Alexander Flint, MD, PhD, is a neurologist and stroke specialist who practices via telemedicine in Northern California. His research focuses on stroke, neurological critical illness, and the use of ‘big data’ and machine learning in medicine. He was a co-founder of the medical imaging startup image32 (acquired by Citrix in 2015) and is currently working on novel technologies that enable machine learning and search directly on securely encoded datasets.
To view Dr. Alexander Flint's publications, visit PubMed