| Risk tool | No of studies | n | Risk of bias | Inconsistency | Indirectness | Imprecision | Mean effect size (95% CI or 95%CI range if >1 study for AUC) | GRADE overall quality |
|---|---|---|---|---|---|---|---|---|
| Mehran risk tool | 11 | 8374 | Very high1 | Very high2 | High3 | Very high4 | Median AUC= 0.780 (0.480-0.912) | VERY LOW |
| Mehran risk tool (cut-off: >5) | 3 | 910 | Very high1 | Low | High3 | Very high5 | Sensitivity= 75.7% (45.3-92.6) | VERY LOW |
| Very high1 | Low | High3 | High6 | Specificity= 73.8% (47.9-89.7) | VERY LOW | |||
| Mehran risk tool (cut-off: >7) | 1 | 644 | Very high7 | NA | High3 | High8 | Sensitivity= 64.1% (52.0-75.0) | VERY LOW |
| Very high7 | NA | High3 | Low | Specificity= 54.9% (51.0-59.0) | VERY LOW | |||
| Mehran risk tool (cut-off: ≥10) | 1 | 301 | Very high7 | NA | High3 | Very high5 | Sensitivity= 64% (44.0-81.0) | VERY LOW |
| Very high7 | NA | High3 | Low | Specificity= 62% (56.0-68.0) | VERY LOW | |||
| Marenzi risk score | 3 | 3920 | Very high9 | High10 | High3 | High11 | Median AUC= 0.57 (range: 0.51-0.83) | VERY LOW |
| Bartholomew risk score | 2 | 1735 | Very high7 | Low | High3 | Very high4 | AUC= 0.59 (0.47-0.72) | VERY LOW |
| Ghani risk score | 2 | 1737 | Very high1 | Low | Low | High12 | AUC= 0.55 (0.41-0.67) | VERY LOW |
| Ando risk score | 2 | 1373 | Very high13 | Very high2 | High3 | High11 | AUC= 0.70 (0.50-0.92) | VERY LOW |
| Gurm (reduced model) risk score | 2 | 21,819 | Very high14 | Very high2 | High3 | High11 | AUC= 0.69 (0.51-0.86) | VERY LOW |
| Inohara risk score | 2 | 2272 | Very high7 | Low | Low | High11 | AUC= 0.705 (0.600-0.770) | VERY LOW |
| Tziakas risk score | 2 | 1447 | Very high13 | Very high2 | High3 | Very high4 | AUC= 0.68 (0.46-0.93) | VERY LOW |
| ACEF score | 2 | 1522 | Very high9 | Low | Low | High11 | AUC= 0.791 (0.656-0.850) | VERY LOW |
| Victor risk score (cut-off: 10%) | 1 | 300 | Very high15 | NA | High3 | High16 | Sensitivity= 92.3% (75-99) | VERY LOW |
| Very high15 | NA | High3 | High6 | Specificity= 82.1% (77-86) | VERY LOW | |||
| GRACE score | 1 | 216 | Very high1 | NA | Low | Low | AUC= 0.828 (0.724-0.932) | LOW |
| GRACE score (cut-off >142) | 1 | 216 | Very high1 | NA | Low | Very high5 | Sensitivity= 81.0% (58.0-95.0) | VERY LOW |
| Very high1 | NA | Low | Low | Specificity= 71.0% (64.0-77.0) | LOW | |||
| GRACE score (cut-off >160) | 1 | 251 | Very high7 | NA | Low | High16 | Sensitivity= 79.1% (64.0-90.0) | VERY LOW |
| Very high7 | NA | Low | Low | Specificity= 61.0% (54.0-68.0) | LOW | |||
| de Ferrari risk score | 1 | 1782 | Very high17 | NA | Low | Low | AUC= 0.838 (0.802-0.874) | LOW |
| CH2DS2-VASc score (cut-off: ≥4) | 1 | 300 | High18 | NA | Low | Low | AUC= 0.81 (0.73-0.90) | MODERATE |
| High18 | NA | Low | High16 | Sensitivity= 90.2% (77.0-97.0) | LOW | |||
| High18 | NA | Low | Low | Specificity= 62.9% (57.0-69.0) | MODERATE | |||
| Gurm (full model) risk score | 1 | 20,572 | Very high19 | NA | High3 | Low | AUC= 0.852 (0.835-0.869) | VERY LOW |
| Zwolle risk score (cut-off: >2) | 1 | 314 | Very high7 | NA | High3 | Low | AUC= 0.85 (0.78-0.92) | VERY LOW |
| Very high7 | NA | High3 | High16 | Sensitivity= 76.3% (68.0-84.0) | VERY LOW | |||
| Very high7 | NA | High3 | High6 | Specificity= 75.4% (66.0-83.0) | VERY LOW | |||
| Lei risk score (cut-off: >129) | 1 | 643 | Very high18 | NA | Low | Low | AUC= 0.78 (0.73-0.83) | LOW |
| Very high18 | NA | Low | High5 | Sensitivity= 81.2% (72.0-88.0) | VERY LOW | |||
| Very high18 | NA | Low | Low | Specificity= 63.3% (58.0-66.0) | LOW | |||
| Liu risk score | 1 | 428 | Very high20 | NA | High3 | High11 | AUC= 0.693 (0.608-0.779) | VERY LOW |
| Liu full risk score | 1 | 1041 | Very high21 | NA | Low | Low | AUC= 0.858 (0.794-0.923) | LOW |
| Liu reduced risk score | 1 | 1041 | Very high21 | NA | Low | Low | AUC= 0.854 (0.796-0.913) | LOW |
| Maioli risk score | 1 | 1247 | High18 | NA | Low | Low | AUC= 0.58 (0.56-0.61) | MODERATE |
| Brown risk score | 1 | 1247 | High18 | NA | Low | High12 | AUC= 0.52 (0.47-0.56) | LOW |
| Tsai risk score | 1 | 1247 | High18 | NA | High3 | High12 | AUC= 0.51 (0.49-0.54) | VERY LOW |
| Caspi risk score | 1 | 1247 | High18 | NA | Low | Low | AUC= 0.53 (0.51-0.56) | MODERATE |
| Liu risk score | 1 | 1247 | High18 | NA | Low | High12 | AUC= 0.52 (0.48-0.57) | LOW |
| Victor risk score | 1 | 1247 | High18 | NA | High3 | Low | AUC= 0.54 (0.50-0.59) | LOW |
| Gao risk score | 1 | 1247 | High18 | NA | High3 | High12 | AUC= 0.49 (0.45-0.53) | VERY LOW |
| Fu risk score | 1 | 1247 | High18 | NA | High3 | High12 | AUC= 0.50 (0.46-0.54) | VERY LOW |
| Chen risk score | 1 | 1247 | High18 | NA | Low | High11 | AUC= 0.48 (0.43-0.52) | LOW |
| McCullough risk score | 1 | 1247 | High18 | NA | High3 | Low | AUC= 0.58 (0.53-0.62) | LOW |
Downgraded by two increments due to very high risk of bias arising from multiple domains of the PROBAST risk of bias tool, most frequently due to unclear definition and assessment of predictors (timing and criteria not specified), unclear interval between predictor and outcome assessment (not specified when predictors were assessed), unclear flow of participants through the study (missing data with no imputation of missing values), inadequate sample size (<100 events) and incomplete analysis reporting (discrimination reported without calibration)
Downgraded by two increments due to substantial differences between the point estimate and 95%CI’s reported in studies examining the same risk prediction tool
Downgraded by one increment due to high levels of concern surrounding the applicability of the risk prediction tool (not all predictors available at the intended time of assessment (prior to contrast administration))
Downgraded by two increments due to the 95%CI overlapping both the upper and lower thresholds for decision making (0.50-0.70)
Downgraded by two increments due to the 95%CI overlapping both the threshold corresponding to ‘low sensitivity’ (60%) and ‘high sensitivity’ (80%)
Downgraded by one increment due to the 95%CI overlapping the threshold corresponding to ‘low specificity’ (80%)
Downgraded by two increments due to very high risk of bias arising from multiple domains of the PROBAST risk of bias tool, namely due to unclear definitions and assessments of predictors (timing and criteria not specified), inadequate sample size (<100 events) and incomplete analysis reporting (discrimination reported without calibration)
Downgraded by one increment due to the 95%CI overlapping the threshold corresponding to ‘low sensitivity’ (60%)
Downgraded by two increments due to very high risk of bias arising from multiple domains of the PROBAST risk of bias tool, most frequently due to inadequate sample size (<100 events) and incomplete analysis reporting (discrimination reported without calibration)
Downgraded by one increment due to considerable differences between the point estimate and 95%CI’s reported in studies examining the same risk prediction tool
Downgraded by one increment due to the 95%CI overlapping the upper threshold for decision making (0.70)
Downgraded by one increment due to the 95%CI overlapping the lower threshold for decision making (0.50)
Downgraded by two increments due to very high risk of bias arising from multiple domains of the PROBAST risk of bias tool, namely due to inadequate sample size (<100 events) and concerns arising from the analysis method (model developed using univariate analysis to identify relevant predictors, unclear if the validation study applied the risk prediction tool as intended)
Downgraded by two increments due to very high risk of bias arising from multiple domains of the PROBAST risk of bias tool, namely due to inadequate sample size (<100 events), unclear definition and assessment of predictors (timing and criteria not specified) and concerns arising from the analysis method (model development study validated the tool using random split sampling and unclear if the external validation study applied the risk prediction tool as intended)
Downgraded by two increments due to very high risk of bias arising from multiple domains of the PROBAST risk of bias tool, namely due to unclear definition and assessment of predictors (timing and criteria not specified), inadequate sample size (<10 events per predictor) and concerns arising from the analysis method (model developed using univariate analysis to identify relevant predictors and random split sampling to validate)
Downgraded by one increment due to the 95%CI overlapping the threshold corresponding to ‘high sensitivity’ (80%)
Downgraded by two increments due to very high risk of bias arising from multiple domains of the PROBAST risk of bias tool, namely due to unclear definition and assessment of predictors (timing and criteria not specified), inadequate sample size (<10 events per predictor) and concerns arising from the analysis method (model developed using univariate analysis to identify relevant predictors)
Downgraded by one increment due to high risk of bias arising from the PROBAST risk of bias tool, namely due to inadequate sample size (<100 events)
Downgraded by two increments due to very high risk of bias arising from multiple domains of the PROBAST risk of bias tool, namely due to unclear definition and assessment of predictors (timing and criteria not specified), inadequate sample size (<10 events per predictor) and concerns arising from the analysis method (model validated using random split sampling)
Downgraded by two increments due to very high risk of bias arising from multiple domains of the PROBAST risk of bias tool, namely due to inadequate sample size (<100 events) and concerns arising from the analysis method (model developed using univariate analysis to identify relevant predictors and random split sampling to validate)
Downgraded by two increments due to very high risk of bias arising from multiple domains of the PROBAST risk of bias tool, namely due to inadequate sample size (<10 events per predictor) and concerns arising from the analysis method (model validated using random split sampling)
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.