Background: Clinical decision rules (CDR), or clinical prediction rules, are tools for clinicians ‘at the bedside’ to assist patient management. Undertaking systematic reviews of CDR in predicting adverse consequences of infection in cancer offered challenges in synthesising such data. Objective: To highlight challenges raised by systematic reviews of the discriminatory ability and predictive accuracy of CDR and serum markers of infection in febrile neutropenia (FNP) in children and young people and posit solutions. Methods: We use the results of two reviews (HTA Registry of Systematic Reviews, CRD32009100453 and CRD32009100485) to identify challenges in study design and synthesis. Markov chain Monte Carlo (MCMC) models using advanced hierarchical statistical modelling taking into account multiple and different thresholds between studies were used for meta-analysis. Results: We found 24 studies describing 17 different CDR in 8388 episodes of FNP. No study compared different approaches and two CDR, producing two and three-level test results, were subject to meta-analysis. 26 studies examined 13 markers in 3585 episodes; four datasets were pooled. A range of methodological problems with the primary studies were identified. These included: small event-per-variable ratios, lack of shrinkage of predictive estimates, failure to examine for non-linear relationships, use of data-driven variable selection and cutpoint determination, premature categorisation of continuous data, lack of examination of missing data, and suboptimal examination of clustered data. Many of the problems could be assisted by the collection of individual participant data (IPD). Depending on the clinical question, a ‘diagnostic’ or ‘prognostic’ approach to synthesis becomes more clinically meaningful. Conclusions: 1. CDR studies contain similar problems as prognostic and diagnostic studies and may be best assessed using IPD. 2. MCMC models can be used to synthesise the information efficiently these are not available. 3. Advancing a philosophy of prediction, rather than ‘diagnostic accuracy’ or ‘prognostics’ may aid clinicians to understand tests better.