PP57 Grading The Quality Of Evidences In HTA Process

Abstract

Introduction:In decision-making processes, health technology assessment (HTA) plays an important role ensuring the adoption of effective technologies and translating scientific evidence into decisions. Bambino Gesù Children's Hospital developed a new method which integrates EUnetHTA Core Model with multi-criteria decision analysis (MCDA) enabling decision makers to make a more informed decision between different alternatives. This approach quantifies assessment parameters, which are defined by literature evidence, or by expert opinion when lacking such evidence. MCDA results (i.e. decision tree of assessment elements, weighting systems and numerical values of technology’ performance) are derived from expert judgement. This means that indicators are weighed by the same weight system; either they are supported by strong literature evidence or otherwise based on expert opinion. The objective of this work is to use the GRADE approach to weight the relevance of each indicator starting from its source of information because different level of evidence should result in different weights.Methods:A GRADE level was associated with each judgement value of performance indicators and a normal probability function was built with the standard deviation inversely proportional to GRADE level to describe the possible dispersion of the judgement due to the different levels of evidence that support each indicator. The higher the GRADE value, the lower the associated standard deviation. A Monte Carlo simulation was carried out to evaluate the expected value of technology’ performance modulated by GRADE level.Results:Four Gaussian distributions were built and associated to four GRADE levels. When an indicator has a low GRADE level, its performance value will vary in a broader way according to the linked Gaussian distribution.Conclusions:This study showed the importance of applying the GRADE system to indicators’ sources of information because this can modify the overall computation of parameter weights and performance, proportionally to their robustness.