Research Project
8601311
DESCRIPTION (provided by applicant): Critical care has excellent measures of severity of illness calibrated to mortality, but severity may be reflected in subsequent morbidity as well survival. A major challenge of critical care outcomes research and applicable to all medical outcomes and quality issues is the development of methods that predict the full range of outcomes from normal through the range of morbidities as well as death. The AIM of this proposal is to develop and validate a predictor of 3 or more outcome states from pediatric intensive care: death, survival one or more states of reduced functional status, and survival with normal or unchanged functional status. Preliminary Studies demonstrate a) the feasibility of the statistical approach and b) the applicability and utility of a new functional status assessment method (Functional Status Score, FSS) developed by the CPCCRN and by this PI for the purpose of this proposal. METHODS: Consecutive patients without exclusion from the participating PICUS will be utilized. Core data will consist of physiological data, diagnoses, age and other demographic information, FSS (pre-admission, PICU discharge, hospital discharge), survival/death (PICU and hospital), therapies affecting functional status, imaging, Outcome prediction for multiple functional states with normal function and death being the extreme will include both simple linear models with the FSS contributing the gradations of outcome, and polychotomous logistic regression analysis for models of 3 or more discrete outcome states. Statistical models will use up to 12 predictor variables including PRISM III score without neurological variables, neurological variables only, pre- ICU care area, operative status, diagnoses (up to 6), age, baseline FSS. Statistical methods will include simple linear regression conceptualizing outcome on a scale of normal to death with worsening functional states in between and polychotomous logistic regression utilizing the FSS to define 2 of more outcome states in addition to death. Sample size estimates based on a 4% mortality rate and a 4% new severe functional status are 5067 but will be re-estimated when units are selected.
