A framework for a decision analysis method that is based on the principle of maximization of the expected utility regarding alternatives of age assessment for unaccompanied asylum seekers is here presented. Using the framework, different methods (dental, knee joint, hand wrist and the method used by The National Board of Forensic Medicine (RMV) that combines the methods for dental och knee joint) for medical age assessment are compared. These methods are further compared with three benchmark alternatives, (i) to trust the age given by the unaccompanied asylum seeker which results that all are considered to be children, (ii) to consider all the unaccompanied asylum seekers as adults and (iii) the absurd alternative to flip a coin to decide who is a child or an adult.
For the decision analysis, assumptions and/or estimates for the age distribution of the unaccompanied refugees are needed and estimates for the different methods regarding how probably it is to be considered an adult given the actual age. The outcome of a child that is incorrectly classified is assumed to give the lowest utility and a correct classification (both children and adult) is assumed to give the highest utility. The utility of the outcome of an adult that is incorrectly classified as a child needs to be quantified.
The age distribution of unaccompanied refugee, considered for age assessment is here assumed to be a combination of two continuous uniform distributions, with the interval 15-18 years (child) and 18-21 years (adult). Two utility models are examined, a discrete model that only consider if the individual is a child or an adult and a continuously linear utility model that consider the age difference from 18 years given an incorrect classification.
The analyzes carried out demonstrates how the framework can be used in practice. Given the assumptions that are made the conclusion is that the alternative that gives the highest expected utility depends on the prevalence (proportion of adults) together with the valuation of the utility for an incorrect classified adult. Regardless of the valuation, when the prevalence is close to 0 all should be considered to be children, when the prevalence increases should be replaced with a method that largely classifies children correct when the prevalence is further increased should be replaced with a method that largely classifies adults correct and finally, when the prevalence is close to 1 all should be considered as adults.