Abstract
This study concerns statistical methods for estimating the health care cost attributable to a disease for a defined patient. Such estimates may be of value for health economic evaluations. Ways of handling incomplete (or censored) cost and survival data are discussed, with application to data on survival and in-patient resource utilization in head and neck cancer. The database was obtained from the national Swedish cancer registry and includes all patients diagnosed with head and neck cancer in Sweden from 1986 to 1996. The main method of estimating the average in-patient costs attributable to head and neck cancer is the Kaplan-Meier sample-average estimator, which takes account of censored survival and cost data. A parametric analogue to the Kaplan-Meier sampleaverage estimator is also presented. Towards the end, alternative methods are discussed, e.g. the possibility of refining the analysis by taking accooot of explanatory ariables in a regression model. As for the results, the analysis presented in the study suggests that the average in-patient cost attributable to head and neck cancer in Sweden is about SEK 260 000 per patient from diagnosis to death (2001 prices).