Aims & Objectives:
1. explore key variables thought to be associated with the difficult-to-sedate child; propose a conceptual model linking those variables in critically ill children
2. assess face and content validity of candidate variables identified in the conceptual model
3. build and test a statistical model describing the difficult-to-sedate child clinical phenotype
Literature review identified the lack of an operational definition of the difficult-to-sedate child and identified factors possibly contributing to the clinical phenotype. These factors were used to develop an initial definition and construct a conceptual model. Expert pediatric critical care clinicians validated the elements of the definition through assessment of face and content validity and proposed additional factors for inclusion in the model via a web-based survey. A refined definition was tested using data from the RESTORE study. Variability in sedation response was characterized using Latent Class Analysis. Classification and Regression Tree methodology was used to identify characteristics of patients at risk for being difficult-to-sedate.
Latent Class Analysis identified a two-class model as the best fit. Need for adjunctive medications, less organ failure, occurrence of inadequate sedation events, and normal cognitive state were indicative of the difficult-to-sedate child latent class. Classification and Regression Tree analysis produced a tree with 9 nodes. The best fitting model classified 18% of children as likely to be difficult-to-sedate.
The conceptual model and operational definition require further investigation and refinement, as well as prospective validation by other investigators. This study suggests that a clinically meaningful population of difficult-to-sedate children requiring mechanical ventilation for a critical illness exists.