Comparative accuracy of risk prediction models for mortality in acute coronary syndrome: A protocol for systematic review and meta analysis.
The accuracy of different risk prediction models must be directly compared using research evidence from each model. This study systematically collected, evaluated and synthesized comparative accuracy data of mortality risk models for acute coronary syndrome (ACS) patients to compare their performance.
An evidence-based approach was used to investigate ACS mortality risk prediction models. First, we searched multiple databases from 2009 to 2024, to identify multivariate predictive models for predicting ACS mortality risk. Included studies were screened, quality-assessed, and data extracted. PROBAST evaluated the risk of bias; heterogeneity was analyzed via MetaDiSc1.4 (I2 statistic). Data analysis used RevMan5.3 and MetaDiSc1.4. Sensitivity (SEN), specificity (SPE), positive/negative likelihood ratios (LR+/LR-), and area under the curve (AUC) of models were calculated for comparison.
A total of 8277 documents were retrieved, and 6 studies were finally included, involving 5 risk prediction models, a total of 24,911 patients with ACS, including 18,443 males (74.04%) and 6468 females (25.96%), with 1637 deaths. The SEN of the global registry of acute coronary events (GRACE) model was 0.78, SPE was 0.76, and AUC was 0.86; the SEN of the thrombolysis in myocardial infarction model was 0.51, SPE was 0.81, and AUC was 0.64; the SEN of the rapid emergency medicine score (REMS) model was 0.78, SPE was 0.46, and AUC was 0.41. The Acute physiology and chronic health evaluation II and REMS2 were reported separately due to non-combinable effect sizes, with SEN 0.77 to 0.95, SPE 0.22 to 0.99, and AUC 0.71-0.92. All 6 studies compared model accuracy. Pooled evidence indicated GRACE (AUC = 0.79) outperformed thrombolysis in myocardial infarction (0.59) and REMS (0.41); APACHE II (0.82) outperformed REMS (0.61) but was slightly inferior to GRACE (0.86).
The GRACE risk prediction model is highly accurate and includes comprehensive clinical research data. It allows medical staff to accurately assess the death risk of ACS patients and effectively reduce their mortality. Therefore, the study suggests that clinical nursing staff use the GRACE risk prediction model to assess the risk of death in patients with ACS.
An evidence-based approach was used to investigate ACS mortality risk prediction models. First, we searched multiple databases from 2009 to 2024, to identify multivariate predictive models for predicting ACS mortality risk. Included studies were screened, quality-assessed, and data extracted. PROBAST evaluated the risk of bias; heterogeneity was analyzed via MetaDiSc1.4 (I2 statistic). Data analysis used RevMan5.3 and MetaDiSc1.4. Sensitivity (SEN), specificity (SPE), positive/negative likelihood ratios (LR+/LR-), and area under the curve (AUC) of models were calculated for comparison.
A total of 8277 documents were retrieved, and 6 studies were finally included, involving 5 risk prediction models, a total of 24,911 patients with ACS, including 18,443 males (74.04%) and 6468 females (25.96%), with 1637 deaths. The SEN of the global registry of acute coronary events (GRACE) model was 0.78, SPE was 0.76, and AUC was 0.86; the SEN of the thrombolysis in myocardial infarction model was 0.51, SPE was 0.81, and AUC was 0.64; the SEN of the rapid emergency medicine score (REMS) model was 0.78, SPE was 0.46, and AUC was 0.41. The Acute physiology and chronic health evaluation II and REMS2 were reported separately due to non-combinable effect sizes, with SEN 0.77 to 0.95, SPE 0.22 to 0.99, and AUC 0.71-0.92. All 6 studies compared model accuracy. Pooled evidence indicated GRACE (AUC = 0.79) outperformed thrombolysis in myocardial infarction (0.59) and REMS (0.41); APACHE II (0.82) outperformed REMS (0.61) but was slightly inferior to GRACE (0.86).
The GRACE risk prediction model is highly accurate and includes comprehensive clinical research data. It allows medical staff to accurately assess the death risk of ACS patients and effectively reduce their mortality. Therefore, the study suggests that clinical nursing staff use the GRACE risk prediction model to assess the risk of death in patients with ACS.