Statistical analysis

Sample characteristics and causes of death of all patients who died (n = 390) were analysed descriptively and were presented as median and quartiles for continuous variables and as frequencies and percentages for categorical variables.

Healthcare utilization was analysed for the 327 patients who had a cause of death and died non-suddenly and non-accidently, as these patients could have had end-of-life care needs and death could have been expected. Healthcare utilization in the last 12 months of life was analysed descriptively and logistic regression analyses were used to examine differences between patients who died due to either a malignant or a cardiovascular cause of death. For contacts with a GP, hospital days, and ED visits, the mean monthly days with an event were reported for every month in the last year of life and modelled using binomial regressions with generalized estimating equations. For the analyses of GP and ED visits, only the days that patients were not hospitalized (starting from the second day of the hospital admission) were taken into account, as patients usually do not have GP contacts or ED visits during a hospital admission. For contacts with CHD physicians, cardiovascular procedures, admissions to the ICU, and specialist palliative care, it was considered inappropriate to calculate the mean monthly days for each item, because of excessive zeros in the data. Therefore, for these types of healthcare utilization, the proportions of patients using this healthcare service at least once per month for the last 12 months of life were reported, together with exact confidence limits for the proportions.