Substance abuse is viewed as one of the main factors (criminogenic needs) to be assessed and targeted in prison. Prison assessments of risk and needs are known to validly predict reoffending. However, there seems to be lacking research in how reliably the individual prisoner’s problems, such as substance abuse, are represented in the assessments. In this study, we compare an independent medical health study (N = 510) with in-prison assessments for the same persons to see whether some of the inmates’ substance abuse disorders were overlooked in prison. We found that sentence plans (257) were in poor agreement with the health study (Kappa 0.315); they recognized only 65 % of all diagnoses. The risk and needs assessments (178) were in closer agreement with the health study, however, alcoholism diagnoses were recognized less accurately (Kappa 0.519) and less frequently (78 %) than drug diagnoses (Kappa 0.627, 87 %). The main factors predicting an assessment of substance abuse risks in prison, analysed through logistic regression were: longer stay in prison and one or more dependence diagnoses. We conclude that, a number of potentially criminogenic dependence problems remain unrecognized since some groups of prisoners are either completely left out from the more thorough instrument, the risk-and-needs assessment, or are not assessed thoroughly enough. This puts prisoners in unequal positions, since all interventions in prison are based on assessments. The study alerts us of the selectiveness of prisoner assessments in practical settings; the unrecognition of problems of shorter sentenced prisoners and prisoners with alcohol dependence.