Influenza-associated mortality has traditionally been estimated as the excess mortality above a baseline of deaths during influenza epidemic periods. Excess mortality estimates are not timely, because national vital statistics data become available after a period of 2-3 years. To develop a method for timely reporting, we used the 121 Cities Surveillance System (121 Cities), maintained at the Centers for Disease Control and Prevention, as an alternative data source. We fit a cyclical regression model to time series of weekly 121 Cities pneumonia and influenza deaths for 1972-1996 to estimate the excess pneumonia and influenza mortality and to compare these figures with national vital statistics estimates for 20 influenza seasons during 1972-1992. Seasonal excess mortality based on 121 Cities correlated well with the national data: for 18 (90%) of 20 seasons, our influenza epidemic severity index category approximated the result based on national vital statistics. We generated preliminary severity categories for the four recent seasons during 1992-1996. We conclude that the 121 Cities Surveillance System can be used for the timely assessment of the severity of future influenza epidemics and pandemics. Timely pneumonia and influenza mortality reporting systems established in sentinel countries worldwide would help alert public health officials and allow prompt prevention and intervention strategies during future influenza epidemics and pandemics.