OBJECTIVE--To suggest an appropriate denominator for use with the statistic "potential years of life lost", in order to facilitate valid comparisons between health authorities. DESIGN--Analysis of OPCS population and mortality data for the period 1987-91. SETTING--Data for Wales were analysed to provide comparisons between the nine district health authorities. SUBJECTS--There were 46,139 males and 30,258 females--that is, all deaths of Welsh residents, aged under 75 years, during the period 1987-91. MAIN RESULTS--The use of age sensitive denominators with the potential years of life lost statistic alters the rank position of Welsh district health authorities. CONCLUSION--It is suggested that potential years of life lost should be expressed as a rate per 100,000 potential years of life in public health common data sets, and elsewhere, to facilitate valid comparisons between health authorities.