OBJECTIVES: It is difficult to accurately assess the impact of influenza on hospital admissions. Most influenza admissions are not coded as such, since the reason for admission is usually an underlying condition exacerbated by influenza. The difficulty of assessing the excess burden is complicated by the nature of commonly associated conditions such as pneumonia, heart failure, and respiratory disease, all of which exhibit similar seasonal variations in numbers of hospitalizations. The objective is to remove the seasonal variation to estimate the fluctuations attributed to influenza during epidemic periods. METHODS: We used a structural time series model that included a stochastic trend and a trigonometric seasonal function. Dummy variables (1 during influenza epidemic, otherwise 0) were used to capture the excess hospitalizations over and above those that occur due to typical seasonal fluctuation. Weekly hospital data on pneumonia and influenza, all other respiratory conditions, and congestive heart failure were modeled for public hospitals in England and Wales. RESULTS: The models explained between 72% and 95% of the variation in hospital admissions for the relevant diagnostic groups (except CHF, explained variance 43%). We found a one-week lag in two models reflecting the time elapsed between diagnosis and admission. Mean numbers of excess admissions over a 10-year period were: high risk 17,857 (range 12,779 - 26,104); elderly 17,856 (10,869 – 28,966); adult 4,820 (3,175 – 7,453). CONCLUSIONS: Estimation of the burden of excess hospitalizations must take account of underlying seasonal variations. Traditional methods of estimation, such as comparing epidemic and non-epidemic periods, are of limited value. Furthermore, because much of the excess hospital burden of influenza can be hidden (admission codes reflect underlying conditions or complications), it is necessary to take into account the variations in a wide range of conditions.