CPES Score Calculator – Composite Pulmonary Embolism Shock Score

What the original FLASH study established

The 2023 analysis included 500 intermediate-risk PE patients selected for FlowTriever mechanical thrombectomy. In the invasive-hemodynamic analysis, 131/384 (34.1%) met the normotensive-shock definition. Observed prevalence ranged from 0% at score 0 to 58.3% at score 6; the reported odds ratio for 6 versus <6 was 5.84 (95% CI 2.00–17.04). These are cohort results, not individual probabilities. [1]

FLASH was industry-sponsored by Inari Medical. Selection for thrombectomy and nonstandardized site biomarker thresholds limit generalizability. All six source components are required here; there is no missing-value imputation. A small subsequent thrombectomy cohort evaluated the invasive normotensive-shock endpoint, not an automatic treatment indication. [1, 2]

Later research has explored other thresholds for different outcomes

Najarro 2024 studied 848 stable acute-PE patients. A 30-day complicated course (death, hemodynamic collapse or recurrent PE) occurred in 63/848 (7.4%). A score >3 identified 78/848 patients: specificity 92.1%, positive predictive value 20.5%, positive likelihood ratio 3.22 and AUC 0.71. At >4, specificity was 98.1%. These study-specific thresholds are not stages and are not the Table 6 layer. [3]

Zhang 2024 evaluated 207 intermediate-risk patients; 29 (14%) had in-hospital death, resuscitated cardiac arrest or hemodynamic decompensation. The ≥3 group had 22% events versus 2.4% at <3 (adjusted hazard ratio 6.48, 95% CI 1.49–28.04). That composite is different from the original invasive cardiac-index endpoint. [4]

Weekes 2026 independently assessed 1,731 emergency-department patients. In-hospital death or clinical deterioration occurred in 193 (11.1%); advanced intervention in 123 (7.1%); 30-day mortality in 124 (7.2%). At ≥3, sensitivity was 72% and specificity 57% for the primary outcome, with AUC 0.68; advanced-intervention AUC was 0.78. These endpoints must not be combined into a personal risk estimate. [7]

Importantly, the 2026 authors describe CPES as a severity framework rather than a formally derived multivariable prognostic model and do not recommend a specific cutoff for clinical use. Their retrospective availability rule allowed at least five components; this calculator does not copy that rule and requires all six. Broader central-clot definitions in later analyses do not replace the original saddle PE component here. [1, 7]

Clinical deterioration overrides a static score

Hypotension, low MAP, elevated lactate, hypoperfusion, acute kidney injury, reduced urine output, altered mental status, low cardiac index or cardiopulmonary failure require clinical assessment independent of CPES. This calculator neither excludes occult shock nor selects ICU/floor placement, PERT referral, thrombolysis, thrombectomy, medicines, dose, monitoring, admission or discharge. [5]

CPES vs Bova, PESI and other tools

Bova uses four weighted predictors, 0–7 points, HR ≥110 and original stages I–III for PE-related complications. CPES uses six one-point predictors, 0–6 points and HR ≥100 for its shock/severity framework. PESI/sPESI address broader short-term, principally mortality-oriented prognosis. No score is converted into another. [5]

Hestia checks 11 outpatient exclusions and cannot be inferred from low CPES. NEWS2 concerns general physiological deterioration; traditional Shock Index is HR/SBP arithmetic. Wells PE, PERC and YEARS concern suspected PE, not this confirmed-PE task. None is calculated inside CPES. [5]