Some meta-analysis and systematic reviews have provided sufficient evidence that the prognostic outcomes of APE can be affected by haemodynamic status, right ventricular dysfunction, myocardial injury, and other clinical features [37–41]. PESI was designed based on some clinical features. Our meta-analysis showed high-risk PESI is a risk factor for all-cause death, PE-related death, and serious adverse events. Pooled sensitivity, specificity, and SROC were used for predicting PE prognosis. An AUC value of 0.5 indicates that the test has no discriminatory ability, whereas an AUC value of 1.0 indicates perfect capability . While to demonstrate excellent accuracy, the AUC should be in the region of 0.97 or above; an AUC of 0.93 to 0.96 is very good; 0.75 to 0.92 is good; but an AUC less than 0.75 has obvious deficiencies in its accuracy . Our meta-analysis showed the overall weighted AUC for all-cause mortality was 0.78 (95% CI: 0.77 to 0.80); and AUC for PE-related mortality was 0.82 (95% CI: 0.75 to 0.89). PESI has good accuracy for predicting PE prognosis. The main reason is that the PESI was derived form multi-center large sample studies with appropriate research methods. In the PESI study, 15,531 discharged patients with PE treated at 186 hospitals were included, on the basis of the β-coefficients of the stepwise logistic regression model, PESI with a point score was generated .
Our meta-analysis showed in sPESI subgroup, the OR of all-cause mortality, PE-related mortality, and serious adverse events was similar to that in PESI subgroup. Most important, in sPESI subgroup, the AUC for predicting all-cause mortality, PE-related mortality, and serious adverse events was also similar to that in PESI subgroup. It has the same accuracy between PESI and sPESI. The sPESI is also derived form logistic regression analysis . In the derivation data set, univariate logistic regression of the original 11 PESI variables led to the removal of variables that did not reach statistical signifi-cance and subsequently produced the simplified PESI that contained the variables of cancer, chronic cardiopulmonary disease, heart rate, systolic blood pressure, and oxyhemoglobin saturation levels. 11 variables in PESI reduced to 6 variables in sPESI. The sPESI does not decrease prognostic accuracy compared with the original PESI, but sPESI is easier to use.
In general, the performance of prognostic models should be assessed in two ways: its discrimination and calibration aspects. Discrimination is the ability of the model to correctly separate the subjects into different groups. Calibration is the degree of correspondence between the estimated probability produced by the model and the actual observed probability . Discrimination is also called predicting accuracy, which can be assessed by sensitivity (true positive rate), and specificity (true negative rate), ROC, and AUC. Our meta-analysis has already proved the PESI has good accuracy for predicting APE outcomes. The calibration is also assessed by the Hosmer-Lemeshow test . But the data of the calibration was not found in the included studies. So we can’t calculate the calibration for predicting prognosis in this meta-analysis. Some prospective studies for accessing PESI predicting calibration can be recommended.
Except we can’t calculated the calibration for predicting prognosis in this meta-analysis, there are some other limitations. First, Lankeit M maybe used same participants in two included studies  and , as well as Singanayagam A’s studies [26, 29], and Jiménez D’s studies [34, 35]. Repeated include will increase the weight of the studies, and maybe result bias. Second, different PE haemodynamic status of patients included in our meta-analysis, some included haemodynamic stable patients, some included haemodynamic stable and unstable patients. But this bias is small, because evaluation of hemodynamic status is included in the PESI and sPESI. Third, the meta-analysis can’t test the prognostic accuracy against other clinical score such as the Geneva prognostic score. Finally, most of the included studies were came from Europe and North America, only a study came from Asia, no study came from Africa, South America and Oceania. Some high-quality studies should be performed in these areas before PESI used worldwide.