Multiple regression models depend on explanatory variables included

EDITOR,—Although Robert P H Wilson and colleagues acknowledge that demographics and morbidity are important influences on variations in prescribing costs at the level of health authorities, they proceed to consider multiple regression models for variation in prescribing at the level of practices that do not allow for either of these as explanatory variables.1

Analysis of patient linked prescribing data has shown that, within practices, the prescribing rate for those aged 55–64 is about six times that for …