The Pulmonary Embolism Severity Index, or PESI score, is a clinical tool that estimates how likely a person diagnosed with a pulmonary embolism (a blood clot in the lungs) is to die within 30 days. It sorts patients into risk classes ranging from very low to very high, and that classification shapes one of the most consequential decisions in emergency medicine: whether someone needs to be admitted to the hospital or can safely go home on blood thinners. The score is widely endorsed in international guidelines, but it has real blind spots, particularly in older adults and people with cancer, that clinicians and patients should understand.
What the Score Actually Measures
The PESI score uses 11 variables pulled from a patient’s demographics, medical history, and vital signs at the time of diagnosis. These are age, sex, history of cancer, heart failure, chronic lung disease, altered mental status, respiratory rate, heart rate, systolic blood pressure, oxygen saturation, and temperature.1PubMed Central. Is the Pulmonary Embolism Severity Index Being Routinely Used in Clinical Practice? Each variable adds a set number of points. Age itself adds points equal to the patient’s age in years (so a 72-year-old starts with 72 points before anything else is counted), and male sex adds 10 points. Conditions like cancer or heart failure add 30 points each. Abnormal vitals, such as a heart rate above 110 or oxygen saturation below 90%, contribute additional points.
The total is then sorted into five risk classes. Class I (fewer than 66 points) and Class II (66 to 85) are considered low risk, with predicted 30-day mortality rates under about 3–4%. Class III (86 to 105) is intermediate risk. Classes IV (106 to 125) and V (above 125) are high risk, where mortality climbs steeply.2PubMed Central. Is the Pulmonary Embolism Severity Index Being Routinely Used in Clinical Practice? For clinical decision-making, Classes I and II are often grouped together as “low risk” and III through V as “high risk.”3PubMed. The validation and reproducibility of the pulmonary embolism severity index
The Simplified PESI
The original PESI works well, but calculating an 11-variable score in a busy emergency department is not always practical. That led to the development of the simplified PESI (sPESI), which collapses the score down to six yes-or-no questions: Is the patient older than 80? Do they have cancer? Do they have heart failure or chronic lung disease? Is their heart rate above 110? Is their systolic blood pressure below 100? Is their oxygen saturation below 90%? One “yes” to any question equals one point. A score of zero means low risk; one or more means higher risk.
In the original validation study, the simplified and original PESI had nearly identical ability to predict 30-day mortality, with both scoring an area under the curve of 0.75.4JAMA Internal Medicine. Simplification of the Pulmonary Embolism Severity Index for Prognostication in Patients With Acute Symptomatic Pulmonary Embolism That made the simplified version attractive for routine use. However, a larger comparison found the picture is more nuanced: the original PESI classified a greater share of patients as low risk (about 41% versus 37% with the simplified version) and had slightly better discriminatory power overall.5PubMed. A comparison of the original and simplified Pulmonary Embolism Severity Index For short-term predictions at 30 days, both perform similarly enough that the simplified version is considered a reasonable trade-off for speed. But for longer-term prediction at one year, the simplified PESI’s accuracy drops below that of the original.6Journal of Thrombosis and Haemostasis. Pulmonary embolism severity index accurately predicts long‐term mortality rate in patients hospitalized for acute pulmonary embolism
The practical difference matters at the margins. In one study, the mortality rate among patients the simplified PESI labeled “low risk” was under 1%, while the original PESI’s low-risk group had a mortality rate of about 4%.7PubMed Central. Comparing three clinical prediction rules for primarily predicting the 30-day mortality of patients with pulmonary embolism The simplified version is more conservative: it classifies fewer people as low risk, but those it does classify that way are at genuinely very low risk. The original PESI is more generous with its low-risk label but accepts a slightly higher mortality rate within that group. Neither approach is wrong; they reflect different philosophies about where to set the safety threshold.
How the Score Guides Treatment Decisions
The most direct consequence of a low PESI or sPESI score is the possibility of going home instead of being admitted to the hospital. A pulmonary embolism sounds terrifying, and many patients assume they need intensive inpatient monitoring. For high-risk patients, that is absolutely true. But for someone in PESI Class I or II, or with an sPESI of zero, research consistently shows that outpatient treatment with anticoagulants is safe.8PubMed. Safety, feasibility and patient reported outcome measures of outpatient treatment of pulmonary embolism In one study, patients with an sPESI of zero had a 30-day mortality rate of 0%, compared with about 5% for those scoring one or higher.9PubMed Central. Selection of Home Treatment and Identification of Low-Risk Patients With Pulmonary Embolism Based on Simplified Pulmonary Embolism Severity Index Score in the Era of Direct Oral Anticoagulants
Accumulated evidence suggests that outcomes in low-risk patients treated at home are at least as good as those treated in the hospital.10PubMed Central. Reducing the hospital burden associated with the treatment of pulmonary embolism This is not just a theoretical finding. Direct oral anticoagulants (the newer blood thinners taken as pills) have made home treatment far more practical than it was in the era of intravenous heparin drips. Still, even among patients with an sPESI of zero, only about 17% of those diagnosed outside the hospital were actually sent home in one recent cohort.11PubMed Central. Selection of Home Treatment and Identification of Low-Risk Patients With Pulmonary Embolism Based on Simplified Pulmonary Embolism Severity Index Score in the Era of Direct Oral Anticoagulants The gap between what the evidence supports and what happens in practice is wide. Physicians may be cautious for good reasons (social factors, patient anxiety, unreliable follow-up), but it also reflects underuse of the score itself.
Interestingly, one randomized trial that gave emergency physicians access to calculated PESI scores did not find a reduction in hospital length of stay compared with standard care. The median stay was eight days in both groups.12PubMed Central. The Clinical Impact of the Pulmonary Embolism Severity Index on the Length of Hospital Stay of Patients with Pulmonary Embolism: A Randomized Controlled Trial Simply providing a number does not change behavior if the clinical culture around discharge decisions remains unchanged. The score is a tool, not a mandate.
PESI Versus Other Decision Tools
The PESI is not the only way to decide whether a pulmonary embolism patient can go home. The Hestia rule takes a different approach entirely: instead of calculating a score from demographics and vitals, it uses a checklist of practical exclusion criteria. Questions include whether the patient is hemodynamically unstable, whether they need supplemental oxygen, whether they need pain medication via IV, or whether there are social or medical reasons they cannot be safely managed at home. If none of the criteria apply, the patient is considered eligible for outpatient treatment.
A large randomized trial (HOME-PE) compared the two strategies head-to-head. The rate of death, recurrent clots, or major bleeding at 30 days was 3.8% in the Hestia group and 3.6% in the sPESI group, and the Hestia rule was found to be noninferior. Both strategies resulted in roughly one-third of low-risk patients being managed at home.13American College of Cardiology. HESTIA Rule vs. Simplified PESI for Home Treatment of Patients With Acute Pulmonary Embolism – HOME-PE A meta-analysis looking at sensitivity for predicting mortality found the sPESI slightly outperformed Hestia: the risk of wrongly classifying a high-risk patient as low risk was about 2 per 1,000 patients with the sPESI, compared with 5 per 1,000 with Hestia.14PubMed Central. The Accuracy of Hestia and Simplified PESI to Predict the Prognosis in Pulmonary Embolism: Systematic Review with Meta-analysis
In practice, some institutions use one and some the other. The Hestia rule has the advantage of capturing practical barriers to outpatient care that no point-based score can detect. The PESI has stronger validation data. Some emergency departments use both in sequence: the PESI to confirm low risk, then Hestia to make sure nothing practical would make discharge unsafe.
Where the Score Struggles
The PESI was derived from general adult populations, and its accuracy drops in specific groups where the underlying assumptions break down. The two most important are older adults and people with active cancer.
Older Adults
Because age itself is the single largest contributor to the PESI score (literally equal to the patient’s age in years), an 80-year-old starts with 80 points before any clinical finding is even considered. Add male sex and a single abnormal vital sign, and they are already in a high-risk class regardless of how well they look. Research on octogenarians confirms this concern: the sPESI classifies most patients over 80 as high risk, giving it poor specificity in this group, while the original PESI performs somewhat better but still loses accuracy compared to younger populations.15PubMed. Too Old for PESI?: Risk Stratification of Octogenarians with Pulmonary Embolism in the Emergency Department One study found that scores not incorporating age, like the Geneva Prognostic Score, identified a larger share of elderly patients as low risk, though with somewhat lower prognostic accuracy.16Journal of Thrombosis and Haemostasis. Prospective comparison of clinical prognostic scores in elder patients with a pulmonary embolism
The upshot is that older adults are over-triaged by PESI. Many patients in their 70s and 80s who could safely be managed at home end up admitted because their age alone pushes them past the risk threshold. Combining scoring systems, particularly PESI with the Bova score (which focuses on markers of right heart strain), may improve accuracy in geriatric patients.17PubMed Central. Predictive Value of Wells, Geneva, Bova, and PESI Scores in Elderly Pulmonary Embolism Patients
Cancer Patients
People with active cancer are at high risk for pulmonary embolism, but their risk of dying from it is tangled up with their risk of dying from the cancer itself. The PESI gives 30 points for a cancer diagnosis, which seems reasonable, but the score still struggles in this population because its other variables do not capture the complexity of cancer-related mortality. In one study, the PESI classification did not significantly predict 30-day mortality in cancer patients, and its discriminatory power was barely better than a coin flip.18PubMed. Prognostic Value of Treatment Setting in Patients With Cancer Having Pulmonary Embolism: Comparison With the Pulmonary Embolism Severity Index
A systematic review of prediction tools for pulmonary embolism in cancer patients found that while most tools (including PESI) were good at identifying patients who would die (sensitivities above 93%), they were much worse at correctly identifying those who would not. Specificities ranged from only 6% to about 54%, meaning many cancer patients who would have been fine were nonetheless flagged as high risk.19Journal of Thrombosis and Haemostasis. Prognostic accuracy of clinical prediction rules for early post-pulmonary embolism all-cause mortality in patients with cancer: a systematic review and meta-analysis Cancer-specific tools like the POMPE-C score were developed to address this gap, but none has yet achieved widespread adoption.
Why Biomarkers and Imaging Matter Alongside the Score
The PESI is a clinical score: it uses only bedside information. It does not incorporate blood tests or imaging. This is by design (simplicity is the point), but it also means the score misses something important: right ventricular dysfunction. When a large clot blocks blood flow through the lungs, the right side of the heart has to work harder, and that strain can lead to rapid deterioration even in a patient whose vitals look reasonable at the moment of scoring.
An individual patient data meta-analysis found that among patients classified as low risk by clinical models, those with right ventricular dysfunction (detected on echocardiography, CT, or by elevated levels of BNP/NT-proBNP) had a short-term death rate of about 1.5%, compared to 0.3% in those without it. For PE-related death specifically, the odds ratio was dramatic.20European Heart Journal. Right ventricle assessment in patients with pulmonary embolism at low risk for death based on clinical models: an individual patient data meta-analysis Echocardiography findings remained an independent predictor of bad outcomes even after accounting for the PESI class, meaning it adds genuinely new information the score cannot capture on its own.21European Respiratory Journal. Echocardiography and pulmonary embolism severity index have independent prognostic roles in pulmonary embolism
Similarly, combining high-sensitivity cardiac troponin (a blood marker of heart muscle injury) with the PESI score improves risk stratification beyond what either provides alone.22PubMed Central. Utility of Combining High-Sensitive Cardiac Troponin I and PESI Score for Risk Management in Patients with Pulmonary Embolism in the Emergency Department In practice, many hospitals now follow a two-step approach: first calculate the PESI or sPESI, then check troponin and imaging in patients near the decision threshold. This layered approach identifies “intermediate-risk” patients: those who look okay on paper but have signs of heart strain that warrant closer monitoring.
One study that looked at pulmonary embolism response teams (PERTs) found that right ventricular dysfunction was a better discriminator between risk categories than the PESI score alone. The PESI score cut-points that distinguished low-risk from submassive PE significantly underestimated observed mortality, suggesting the score’s predictions do not fully capture hemodynamic reality.23PubMed Central. Right Ventricular Dysfunction is Superior and Sufficient for Risk Stratification by a Pulmonary Embolism Response Team
Automating the Score in Electronic Health Records
If the PESI is meant to be used routinely, calculating it by hand is a barrier. Several hospitals have built automated PESI calculators into their electronic health records (EHRs) that pull the 11 variables from existing patient data and generate a score without requiring a clinician to enter anything manually. The idea is appealing, but execution is tricky.
A study evaluating one such automated system found it exactly matched the manually calculated PESI in about 79% of cases. More reassuringly, it correctly sorted patients into the right risk group (low versus high) 95% of the time.24JMIR Medical Informatics. Performance of an Electronic Health Record–Based Automated Pulmonary Embolism Severity Index Score Calculator: Cohort Study in the Emergency Department The errors were not random: accuracy dropped for patients who were new to the health system and had no prior encounters, because the algorithm could not reliably determine their medical history. Chronic conditions like cancer or heart failure were the variables most often missed or misclassified. For patients with established records in the system, the automated score performed well. But for someone showing up at an emergency department for the first time, manual calculation remains more reliable.
Machine Learning Approaches
Researchers have begun testing whether machine learning models can outperform the PESI. These models can ingest far more variables, including lab results, imaging features, and complex interactions between risk factors, without requiring a human to decide which variables matter most. A systematic review and meta-analysis concluded that machine learning tools show promise in enhancing PE risk assessment.25PubMed Central. Advancing Mortality Prediction in Pulmonary Embolism Using Machine Learning Algorithms—Systematic Review and Meta‐Analysis
In one cohort study, a machine learning risk score predicted 30-day mortality with an area under the curve of 0.71, compared with 0.65 for the sPESI and 0.64 for the original PESI.26PubMed. Predicting acute and long-term mortality in a cohort of pulmonary embolism patients using machine learning Those are modest improvements. The machine learning model’s advantage was more pronounced for long-term mortality prediction, where traditional scores have always been weaker. Whether these models will eventually replace the PESI in everyday practice depends on whether they can be embedded seamlessly into clinical workflows. A bedside score that a doctor can calculate in 30 seconds has a built-in advantage over an algorithm that requires structured data extraction and real-time processing. For now, the PESI remains the workhorse, but its ceiling is visible, and more data-hungry models are likely to carve out roles in complex cases or integrated decision-support systems.
Pregnancy and Other Unstudied Populations
The PESI was developed and validated almost entirely in non-pregnant adults. Pregnancy creates a unique physiological situation: heart rate naturally increases, blood volume expands, and the risk factors for clotting are different from those in the general population. Pregnant and postpartum patients were not included in the derivation cohorts for either the original or simplified PESI, so the score has no validated role in this group. Most clinical guidelines recommend against using PESI to guide management in pregnancy and instead rely on multidisciplinary teams and case-by-case assessment.
Similarly, pediatric patients fall outside the score’s scope entirely. Children with pulmonary embolism are rare, and the age-based weighting system makes no sense in a 10-year-old. For these populations, the absence of validated tools is itself a clinical challenge, and it underscores that the PESI is a product of the data it was trained on: predominantly middle-aged and older adults presenting to emergency departments in North America and Europe.