Calc Function

    • Calcs that help predict probability of a diseaseDiagnosis
    • Subcategory of 'Diagnosis' designed to be very sensitiveRule Out
    • Disease is diagnosed: prognosticate to guide treatmentPrognosis
    • Numerical inputs and outputsFormula
    • Med treatment and moreTreatment
    • Suggested protocolsAlgorithm





    Chief Complaint


    Organ System


    Patent Pending

    POMPE-C Tool for Pulmonary Embolism Mortality

    Predicts mortality for cancer patients with PE.
    When to Use
    Why Use

    In the setting of a patient with known cancer who has been diagnosed with a pulmonary embolism, the POMPE-C can be used to estimate their mortality.

    The POMPE-C Prediction Tool was risk stratifies patients diagnosed with pulmonary embolism and have active cancer.

    • The POMPE-C prediction tool predicts mortality directly (at 30 days), not a risk category of mortality.
    • Active cancer has been shown to be a risk factor for PE, but less so for cancer in remission.
    • The POMPE-C has not been validated for incidental PE (PE found on imaging but without symptoms), though a small sample was identified and found to be low-risk overall.
    • The POMPE-C may be more accurate in predicting outcome than the PESI Score in the setting of patients with active cancer.
    • The POMPE-C has not yet been validated for selecting patients for outpatient management (though there is potential for future research).
    • Risk stratification is a vital part of the evaluation of a patient diagnosed with PE.
    • Most risk-stratification tools use cancer as a prognostic indicator, but no risk stratification tool exists for those patients with known cancer.
    • Those patients that are determined to be low risk (score ≤5%) may warrant outpatient management (though this has not been validated).


    Please fill out required fields.

    Next Steps
    Creator Insights


    • The POMPE-C score is a well designed and validated prediction rule for patients with known cancer and new PE. Larger prospective validation studies are required before it can be utilized to guide disposition and treatment.
    • The purpose of this study was to set the groundwork for tools with aid in the management of cancer patients found to have incidental PE, as well as those diagnosed with PE who are low risk.
    • Active vs. Inactive cancer may confusing in certain settings, however even inactive had good prediction in the POMPE-C model. (ROC AUC analysis was still 0.85 for inactive cancer with wide confidence intervals).


    • For those patients with a POMPE-C risk of mortality of of ≤5%:
      • Overall risk of mortality is low. No patients in this group died in the study.
      • Consider this group for outpatient management and anticoagulation therapy.
    • For those patients with a POMPE-C mortality estimate >5%:
      • Overall of mortality in this group was not low, but with a posterior probability of 0/50 (95% CI: 0-7%).
      • 106 patients in this group survived >30 days.
      • Consider this group for admission and management.
    • For those patient with a POMPE-C score of >50%:
      • Overall risk of mortality was very high with a posterior probability of 73% (95%CI: 46-95%).
      • Consider this group admission and goals of care discussion.

    Critical Actions

    • No calculator should trump clinical gestalt. A clinically high risk patient should be managed as such, regardless of POMPE-C score.
    • The POMPE-C is only meant for patients with cancer and for risk stratification after the non-incidental diagnosis of PE has been made.
    • Additional pathology and comorbidities should not be overlooked in the setting of a low POMPE-C score.
    • Never delay resuscitative efforts for risk stratification.


    Probability (P) of death using the POMPE-C formula:

    P = 100*(1 - 1 / (1 + Exp(3.718 + 1.551*DNR + 0.799*RespDistress + 0.734*Leg Swelling + 1.473*AMS + 1.027*HR + 0.044*RR + (-0.0636*SAT) + (-0.012*Body Weight))))

    Evidence Appraisal


    • POMPE-C was derived from secondary analysis of a multicenter registry data from 22 emergency departments in the US from 2005 to 2008.
    • Multiple patient variables were collected including lab, ekg and radiology studies. Patients with cancer were also qualified as active cancer, metastatic malignancy or inactive cancer.
    • Incidental findings of filling defects on staging CT scans were excluded.
    • 24 potential variables were compared to 30-day mortality, first with a classification and regression technique (CART) analysis and then by a bivariate analysis (t-test or chi-square). Variables that met statistical standards to be included then underwent ROC curve analysis.
      • 8 variables were identified as predictors:
    Variable Odds Ratio (95%CI)
    Weight 0.99 (0.98-1)
    Highest RR 1.05 (1.01-1.09)
    SaO2 0.94 (0.89-0.99)
    HR > 99 2.8 (1.38-5.64)
    AMS 4.36 (1.47-13)
    Resp. Distress 2.22 (1.12-4.43)
    DNR 4.72 (1.75-12.76)
    Unilateral Limb Swelling 2.08 (1-4.36)
    • A second regression and bivariate analysis was then performed for the included predictors for death at 30 days, with a Pearson’s goodness of fit P value of 0.94 and McFadden’s R-squared of 0.21.
    • Within the derivation cohort 1880 patients had a diagnosis of PE, with 408 having active cancer and 108 having inactive cancer. 51 (12.5%) of the active cancer patients died of any cause within 30 days.
    • AUC under the ROC for each group:
      • Active Cancer 0.84 (95% CI: 0.78 to 0.89)
      • All PE patients 0.8 (95% CI: 0.76 to 0.85)
      • Inactive cancer 0.85 (95% CI: 0.66 to 1.0)
    • PESI score performed well for identifying low risk patients from the derivation cohort, but when applied to the 408 patients with active cancer had a reduced accuracy with an AUC of 0.68 (compared to 0.79 for the entire cohort).


    • Validation was performed prospectively in the US, New Zealand and Europe by research assistants in real time on patients with active cancer. Percentage mortality estimate was calculated.
    • 14,121 patients were identified. 202 had active cancer, but 20 did not complete the enrollment process or were missing data resulting in 182 patients that were included in the final cohort of the validation study.
      • 27 (15%) died within 30 days of the study.
      • AUC of the ROC was 0.86 (95% CI: 0.78 to 0.93).
    • From the logistic equation, cutoffs were selected that predicted very low and very high 30-day mortality probability.
      • A cutoff of 5% was selected as a low probability of death.
        • 50 patients (27%) had a POMPE-C estimate ≤5% with no deaths within 30 days.
      • A cutoff of 50% was selected as a high probability of death.
        • 13 patients (7%) had a POMPE-C estimate >50% and 10 (77%) died.
    • A subanalysis of patients incidentally diagnosed with PE in the emergency department was performed.
      • 16 patients were identified that were not discharged.
        • One died within 30 days, with a POMPE-C of 46%.
        • 15 lived, and 9 had a POMPE-C <5%.
      • 19 other patients were discharged from the emergency department.
        • There were no deaths
        • Mean POPME-C score was 6±8% and 14 had scores ≤5%.


    Dr. Jeffrey Kline

    About the Creator

    Jeffrey Kline, MD, is a professor of emergency medicine and physiology and the vice chair of research at Indiana University. Among other research, he has conducted clinical studies using breath-based methods to diagnose and assess the severity of PE. He co-founded BreathQuant Medical Systems Inc to advance practical applications of 16 patents for medical devices. Dr. Kline has published over 50 manuscripts in the area of PE.

    To view Dr. Jeffrey Kline's publications, visit PubMed

    Content Contributors
    About the Creator
    Dr. Jeffrey Kline
    Content Contributors