Pulmonary Embolism Severity Index (PESI)

Clinical background

The Pulmonary Embolism Severity Index (PESI) was developed to make objective a risk assessment that otherwise rests on clinical intuition: which patients with confirmed acute pulmonary embolism have a 30-day mortality low enough that hospital care is unnecessary, and which require admission or even intensified treatment? The instrument is not a diagnostic tool but a prognostic risk stratification used once the diagnosis has been made. The decision matters clinically because the majority of patients with pulmonary embolism are haemodynamically stable, and a substantial proportion of them can be managed as outpatients if the risk of early death and complications is sufficiently low.

Calculating the Pulmonary Embolism Severity Index

PESI is an additive point model in which the patient's age in whole years forms the base score, after which a further ten binary variables are added with varying weights:

$$\text{PESI} = \text{age} + 10 \times [\text{male}] + 30 \times [\text{previous cancer}] + 10 \times [\text{previous heart failure}] + 10 \times [\text{chronic lung disease}] + 20 \times [\text{HR} \geq 110] + 30 \times [\text{SBP} < 100] + 20 \times [\text{RR} \geq 30] + 20 \times [\text{temp} < 36] + 60 \times [\text{altered mental status}] + 20 \times [\text{SpO}_2 < 90]$$

The derivation cohort consisted of 15,531 inpatients with pulmonary embolism discharged from 186 hospitals in Pennsylvania, USA, divided into a derivation group (67%) and an internal validation group (33%) [1]. The primary outcome was 30-day mortality. The model was derived by logistic regression and stratifies patients into five severity classes. External validation was performed in 221 inpatients from Switzerland and France [1].

The five classes and their 30-day mortality across the derivation and validation cohorts:

In the derivation study, the combined proportion of in-hospital deaths and non-fatal complications was ≤1.1% in class I and ≤1.9% in class II [1].

Interpretation in practice

The main clinical value of the score lies in identifying patients in classes I and II, in whom mortality and the risk of complications are low enough that outpatient treatment can be considered. For patients in class I (≤65 points) the 30-day mortality is at most 1.6% and the combined risk of death and non-fatal complications is below 1.2%. These patients are prime candidates for home treatment, provided that adequate anticoagulation can be assured, that the patient has access to follow-up, and that no other factors (pain, social circumstances, bleeding risk) argue against it.

Class II (66 to 85 points) carries a somewhat higher mortality (up to 3.5%) but still a low complication rate. Outpatient treatment may be appropriate for selected patients here, but the decision requires individual assessment. Many clinicians opt for a short period of inpatient observation in this group.

Classes III to V should be managed as inpatients. For classes III and IV, either a general ward or a monitored unit is appropriate depending on haemodynamic status, whereas class V (>125 points) carries a mortality of up to 24.5% and should be considered for intensive care and possible escalation of treatment, including assessment for thrombolysis in the event of haemodynamic instability.

Validation and performance

In a systematic review and meta-analysis of prognostic models in acute pulmonary embolism comprising 71 studies and 44,298 patients, PESI was the most extensively validated model [2]. Pooled 30-day mortality based on nine validation studies was 2.3% (95% CI 1.7 to 2.9) in the low-risk group and 11.4% (95% CI 9.9 to 13.1) in the high-risk group [2]. The review noted that PESI has also demonstrated clinical utility in an impact study, that is, a study in which the model was actually used to guide management [2].

A simplified version, sPESI, was later developed with six variables scored equally (1 point each) instead of 11 with different weights. In a systematic review comparing sPESI and the Hestia criteria across three studies with a total of 1,608 patients, sPESI had a sensitivity of 0.972 (95% CI 0.917 to 0.991) and a negative predictive value of between 99.0 and 99.4% for 30-day mortality [3]. Specificity was low, however, at 0.269 (95% CI 0.209 to 0.338), meaning that sPESI classifies many patients as high risk who in fact survive [3]. This is an important difference from the original PESI, whose finer scoring gives better discrimination across the whole risk range.

In a comparison with the 2019 ESC risk stratification algorithm, studied in 419 normotensive patients aged ≥65 years, the ESC algorithm placed more patients in higher risk groups than PESI without improving the prediction of short-term outcomes [4]. PESI retained stronger discrimination for 30-day mortality than the ESC algorithm, although the difference did not reach statistical significance [4].

Limitations

PESI applies to patients with confirmed acute pulmonary embolism. It is not a tool for assessing the probability of pulmonary embolism in a patient with a suspected diagnosis, and it should not be used before diagnostic confirmation.

The model was derived in inpatients in Pennsylvania, a population that may differ from contemporary European cohorts in age distribution, comorbidity and treatment practice. The external validation in the original study comprised only 221 patients from Switzerland and France, a relatively small cohort [1].

PESI does not take account of right ventricular function or biomarkers (troponin, NT-proBNP), which are central to the 2019 ESC risk stratification. A patient with a low PESI but demonstrated right ventricular strain may be at increased risk that the score does not capture. In clinical practice, PESI is therefore often supplemented by imaging (echocardiography or CT angiography to assess the right ventricle) and biomarkers, particularly for patients on the border between low and intermediate risk.

Nor does the instrument capture bleeding risk, which is decisive for whether outpatient treatment is safe. A patient with a low PESI but a high bleeding risk (for example recent surgery, thrombocytopenia or active gastrointestinal bleeding) should not be discharged whatever the score.

Age accounts for a large part of the score, which means that older patients automatically fall into higher classes even without other risk factors. This may lead to otherwise healthy older patients being over-treated with inpatient care, and to the risk in young patients with objectively serious embolism being underestimated if they have no comorbidity.

References

1. Aujesky D, Obrosky DS, Stone RA et al. Derivation and validation of a prognostic model for pulmonary embolism. Am J Respir Crit Care Med 2005;172(8):1041–6. PMID: 16020800
2. Elias A, Mallett S, Daoud-Elias M et al. Prognostic models in acute pulmonary embolism: a systematic review and meta-analysis. BMJ Open 2016;6(4):e010324. PMID: 27130162
3. Palas M, Silva BV, Jorge C et al. The accuracy of Hestia and simplified PESI to predict the prognosis in pulmonary embolism: systematic review with meta-analysis. TH Open 2022;6(4):e347–e353. PMID: 36452203
4. Moor J, Baumgartner C, Méan M et al. Validation of the 2019 European Society of Cardiology risk stratification algorithm for pulmonary embolism in normotensive elderly patients. Thromb Haemost 2021;121(12):1660–1667. PMID: 33823559