Aim: To create a classification system for severe, rare, and progressive genetic conditions for use in research reporting. Method: A modified Delphi consensus technique was used to create and reach agreement on a new system of condition categories. Interrater reliability was tested via two rounds of an online survey whereby physicians classified a subset of conditions using our novel system. Overall percentage agreement and agreement above chance were calculated using Fleiss’ kappa (κ). Results: Eleven physicians completed the first Delphi, with an overall agreement of 76.4%, the κ value was 0.57 (95% confidence interval 0.51–0.63), indicating moderate agreement (0.41–0.60) above chance. Based on the first survey several categories were described in more detail. The second survey confirmed a classification system with 12 categories, with an overall percentage agreement among the participants of 82.6%. The overall mean κ value was 0.71 (95% confidence interval 0.65–0.77), indicating substantial agreement (0.61–0.80). Interpretation: Our new system was useful in categorizing a broad range of rare childhood diseases and may be applicable to other rare disease studies; further validation in larger cohorts is required. What this paper adds: This novel 12-category classification system can be used in research reporting in rare and progressive genetic conditions.