Skip to main content

Table 1 (abstract O11). Summary of design ranking in different methodological situations

From: Methods for Evaluation of medical prediction Models, Tests And Biomarkers (MEMTAB) 2018 Symposium

  Stratification design Two-phase design Case-control design Nested case-control design Comprehensive diagnostic study design Comparing two tests without verification of double negatives
Methodological situation:
Population
  Different strata in patient population Only option      
Index test
  Stored data available on index test     Only option   
  Costly or invasive index test Second option   Third option First option   
  Two index tests Third option Second option     First option
  Multiple index tests      Only option  
Reference standard
  Costly or invasive reference standard Third option First option     Second option
Target condition
  Extremely low prevalence (<0.001%)    Only option    
  Multiple target conditions      Only option  
Risk of Bias QUADAS-2 Higher risk of bias if there is more difference between prevalence strata Higher risk of bias if the selection of reference standard is not random Higher risk of bias if cases or controls do not represent the same population Higher risk of bias if the selection of index test is not random Higher risk of bias if there is no reference panel to determine the diagnosis Higher risk of bias if both index tests are not blinded from each other
  1. If there were more than one design applicable, the first three options were ranked on suitability based on sample size and risk of bias. Other designs might be possible as well in specific situations.