1 2 Evidence based bioethics 3 4 4 5 The use of administrative databases 6 An important principle of administrative databases is that they don’t follow patients but money. As soon as money is at stake, administrative registers become highly reliable to count processes amenable to financing, and loss to follow-up is close to negligible. Eligibility for benefits may change over periods, but the fact of being reimbursed is coded extremely well. One of the strong characteristics of administrative databases is their completeness, even over very long periods. In the article of Gravseth et al., early disability pension registered in 1988–2003 is a highly reliable proxy of definitive expulsion from paid work. Looking back to the medical birth registry in 1967–1976 and a limited number of parental characteristics between ages 0 and 18 of the child, the authors have explored the consequences of a limited numbers of determinants in youth for later expulsion out of work. Shaping the life course 7 Moving to evidence based policy Gravseth et al. show, by using linked administrative databases, that causes of early disability in young adulthood are rooted in childhood and in parental disadvantage. A consequence is that interventions targeting the adults at risk will likely add more to the problems than to the solutions. The horns of the dilemma to give or to withhold a disability pension are terribly sharp. An early disability pension is a social death sentence, crippling the individual by removing incentives to seek work and declaring him or her definitively incurable. But those who are effectively crippled won’t be helped by the misery of repeated frustrations in a labour market in which they cannot compete. Studies carefully designed to identify determinants of irreversible disability are needed, inclusive randomised controlled experiments evaluating pension decisions as important health care interventions. 8