Purpose: To propose a new method for calculating the overall utility for longitudinal data, which takes into consideration the changes in utility over time and meantime adjusts overall utility considering distinction between improved and non-improved patients.

Methods: A simulation design was formed to show the accuracy level of new estimator in order to distinguish between improved and non-improved patients, compared to the existing estimators (mean, median, AUC, difference). 10.000 utility values were generated for 12 time points for each estimator. Area under Receiver Operating Characteristic Curve was used to show how accurately these estimates differentiate in improved and non-improved groups for all utility values and utility values for each time point acquired from simulation study.

Results: A new simple nonparametric estimator, overall utility gained  is introduced. Simulation results show that the accuracy of  is higher than other nonparametric estimators to differentiate improved and non-improved groups for longitudinal data. Meanwhile, it is more accurate to consider trend of utilities.

Conclusions: Even though proposed estimator is more accurate than existing non-parametric estimators, it is still not a gold standard to measure overall gained utility. So further research and validation of this new estimator are needed.

utility, longitudinal data, overall utility, decision analysis.