Table 3

Methodological characteristics of included studies

CharacteristicsNumber (%) or median (IQR)
Sample size (n)165 (103–348)
Death events (n)35 (23–72)
Multicentre study
 Yes9 (28.1)
 No23 (71.9)
Epidemiological design
 Prospective cohort13 (40.6)
 Retrospective cohort19 (59.4)
Data sources
 Cohort study5 (15.6)
 EMR data22 (68.8)
 Registry5 (15.6)
Did the study clearly describe inclusion/exclusion criteria for participants?
 Yes31 (96.9)
 No1 (3.1)
Consistent definition/diagnostic criteria of predictors used in all participants
 Yes32 (100.0)
 No0 (0)
Consistent measurement of predictors used in all participants
 Yes32 (100.0)
 No0 (0)
Consistent definition/diagnostic criteria of outcomes used in all participants
 Yes31 (96.9)
 No1 (3.1)
Consistent measurement of outcomes used in all participants
 Yes31 (96.9)
 No1 (3.1)
Were all enroled participants included in the analysis?
 Yes22 (68.8)
 No10 (31.2)
Was missing outcome data reported, and the methods for handling missing outcome
 Yes, complete-case analysis1 (3.1)
 No30 (93.8)
 Not reported1 (3.1)
Was any missing predictor data reported, and the methods for handling missing predictor
 Yes, complete-case analysis5 (15.6)
 No1 (3.1)
 Not reported26 (81.3)
Prognostic factors (n=18) prediction models
Number of outcomes/events in relation to the number of predictors for assessing prognostic factors (EPVs)
 <101 (5.6)
 10–208 (44.4)
 ≥209 (50.0)
Model structure used in the study
 Logistic regression11 (61.1)
 Cox regression5 (27.8)
 ROC analyses (not report regression)2 (11.1)
Relevant model performance measures evaluated for addressing prognostic factors
 AUC2 (11.1)
 AUC, sensitivity, specificity15 (83.3)
 Sensitivity, specificity1 (5.6)
Prediction models (n=14)
Number of outcomes/events in relation to the number of predictors in multivariable analysis (EPVs)
 <103 (21.4)
 10–208 (57.1)
 ≥203 (21.4)
Model structure used in the study
 Logistic regression10 (71.4)
 Cox regression1 (7.1)
 ROC analyses (not report regression)1 (7.1)
 Logistic regression and support vector machines1 (7.1)
 Logistic regression and neural networks1 (7.1)
Relevant model performance measures evaluated for addressing prediction models
 AUC, p value of Hosmer-Lemeshow test5 (35.7)
 AUC4 (28.6)
 AUC, sensitivity, specificity2 (14.3)
 P value of Hosmer-Lemeshow test1 (7.1)
 Expected and observed1 (7.1)
 Sensitivity, specificity1 (7.1)
Develop prediction models (n=11)
Statistical method for selecting predictors during addressing prediction models
 Univariate analysis of predictors by P value3 (27.3)
 Univariate analysis of predictors by p value and other specific predictors3 (27.3)
 Stepwise selection2 (18.1)
 Not reported3 (27.3)
Handling the predictors for addressing prediction models
 Continuous predictor was transformed into categories4 (36.4)
 Not reported7 (63.6)
  • .EMRs, electronic medical records; EPV, events per variable.