[New method of epidemiological surveillance of influenza mortality in the USA]

This paper describes a new method of forecasting the expected number of pneumonia and influenza (P and I) deaths on the basis of time-series analysis of historical mortality data. The new method is different in three aspects from the regression method developed by Serfling and used by the Centres for Disease Control (CDC) since 1963 in that it uses the ratio of P and I deaths to deaths from all causes in lieu of the total number of P and I deaths in forecasting the expected numbers. Also, instead of deleting epidemic periods in the forecast, these weeks are replaced by expected numbers. Finally, forecasting is done with a time-series model instead of the regression model.