[Hospital standardized mortality ratios: a topsy-turvy world].

Hospital standardized mortality ratios (HSMR) are meant to compare mortality risks between hospitals. There is, however, much discussion about the value of the HSMR because incomplete adjustment for case mix is almost inevitable and because the data required to estimate HSMR are often unreliable. There is also a further problem with HSMR, which hampers its interpretation: the starting point for the HSMR is a comparison between the mortality outcomes of more than 50 common diagnoses, not between one or two well-defined treatments for one disease. The number of factors that can influence mortality risk in a hospital for these 50 diagnoses is enormous. There are many potential explanations for an increased (or decreased) HSMR, meaning that even if an increased HSMR really does point towards an increased mortality risk, this provides no real indication of the underlying cause.