Table 1

Identified problems of diagnostic research in febrile children and proposed solutions

Identified problemProposed solution
Patient selection
Patient selection based on age and temperature limitsInclude age or temperature as predictors
Selection and definition of predictors
Heterogeneity of selected predictorsUse a defined set of clinical predictors
Use objective clinical predictors
Broad validation
Clinical valueStepwise analysis of patient history and examination, with evaluation of the added value of diagnostic tests (new markers)
Use rule out symptoms (‘green flags’)
Diagnostic value of ‘new’ predictors and ‘time course’Parental concern
Diagnostic value of ‘safety-netting’
Lack of reference test to confirm bacterial originPragmatic clinical definitions or composite reference test
Include follow up time in the definition of outcome
The changing definition of serious infection/low prevalence of serious outcomeConsider other important outcomes such as need for referral, hospitalisation
Adjust outcome definition to setting
Focus on ruling out versus ruling in serious infections
Heterogeneous outcomes/intermediate outcomes (severity or probability)Polytomous modelling
Generalisability to different clinical settings
Setting with differences in prior risk/(geographic or temporal) changes in case mixStudies in low prevalence settings (primary care) are required
Continuous validation studies
Mismatch of the diagnostic research approach to the clinical diagnostic processInclude repeated assessment of children
Include reassessment plan (safety-netting) in the implementation of decision rule
Lack of impact of the prediction rule on patient outcome/lack of implementation of rule in routine practiceImpact analysis
Implementation strategies