Petter entered the field of forensic statistics, in particular DNA relationship analysis, already in the 1990s but it is not because of that he has become involved in the issue. Instead, it was a colleague at Mathematical Sciences who had a friend that was a trustee for a refugee, and who asked if the statistical argument used in an asylum rejection was indeed scientific. Petter wrote a statement about this and ultimately the person got a residence permit, but of course it is difficult to know if the statement contributed in any way. One and a half year after this, that is last autumn, there was a wave of administrative messages about stricter age assessments, and after a short delay Petter received a corresponding wave of statistically dubious decisions.
– The individual cases I have seen is of course only those where there is an appeal, but common to them was a strange use of statistics. Obviously, it is sad for those people that suffer from a bad use of science, but I also feel a responsibility for statistical science itself so that it does not get a bad reputation if it appears to determine what cannot be supported.
Petter wrote some further statements, about how the medical observations had been used when making decisions, and acquainted himself with literature that existed on the subject. Since last spring, the Swedish Migration Agency demands that medical age assessment of people seeking asylum is performed by the National Board of Forensic Medicine. They use X-rays from wisdom-teeth and NMR scans from knee joints, and in many cases the results are contradictory, that is, one has reached final stage but not the other, and vice versa. Knee joints have not been researched upon to any large extent, and there is great uncertainty about how to interpret these results. The method is being investigated by the National Board of Health and Welfare with the help of people where the age is known, it is to be ready in November, but already results from the method are used. Petter finds this strange – as well as the fact that not results from wrists are used instead, where there are more studies.
There are on the other hand some research around wisdom-teeth, for example a large study in the USA in the 1990s, where traditionally you take a sample from a population, compare the dental status and known age, and describe the age range for a given dental status. A physical marker is chosen that changes a lot at about the age of 18 and which relate to the development stage of the roots of the wisdom-teeth. You observe this age marker for a person seeking asylum, and then use the corresponding spread in the studied population to predict the age of the person. One problem with this is that the individual must be considered to be a sample of that population, and different ethnic groups have different genetics and maturing conditions.
– Another problem is that there is a dependency between the reason that these individuals are age assessed and the variable you want to predict, you suspect that they are of a certain age. This means that you cannot directly use the probabilities you get. In all cases that I have looked upon there is also other information, such as family relationships and psychosocial statements, but the administrators at the Migration Agency do not know how to handle this. They ignore the uncertainty, and the medical age assessments become a too large part of the final decision.
What Petter now wants to do is to evaluate the method of the National Board of Forensic Medicine, to try to estimate how large the errors are statistically, and if the decision-making procedure is optimal. The Ethical Vetting Board has asked him to supplement information on how the personal integrity is preserved and about the competence for assess X-ray sheets, but Petter will only evaluate if the statistics are well done or not, so this information is not really relevant. In the long run he would like to introduce a methodology that, based on a Bayesian decision-making theory framework, may weigh all available information of different types in order to make a decision, in a manner that is reasonable and scientifically feasible also for an administrator at the Migration Agency, and as far as possible minimize the uncertainty and get rid of arbitrariness.
– It is about the well-known problem of optimal decisions under uncertainty and this can, as so much else, be mathematized. Sweden should not go out and say that children are allowed to stay in the country at the same time as people are possibly expelled on loose or incorrect grounds, it is a matter of honesty. Mathematically this can be a simple problem, but it is very political and a real challenge to get an audience for the ideas. I still receive lots of e-mails about decisions that have been made, but I do not have time for them and I am not sure my statements are being used in any case. Instead, I try to influence the system through research.
Text: Setta Aspström
Pictures: NMR scan of knee joint, X-ray of teeth, private