A Bayesian simulation model has been applied to a database developed for split liver transplantation on two adult recipients (SLT A/A) in the context of a macroregional project funded by the Italian Ministry of Health. The model was entered within Bayesian inference Using Gibbs Sampling (WinBUGS), a free software for Bayesian analysis of complex statistical models using Markov chain Monte Carlo techniques developed by the MRC Biostatistics Unit Cambridge jointly with the Imperial College School of Medicine at St Mary's, London. The model was built by using data entry performed from January 1, 2005 to August 5, 2005. In that period, 20 potential donors suitable for the SLT A/A procedure were entered into the database. We only selected the continuous and dichotomous donor-related variables (DRV, n = 62) for which almost one data entry procedure. The model assumed that a database user learned during data entry procedures for each donor, and that the probability of a successful input may depend on the number of previous errors and corrections. After binary transformation of the DRV (value 0 for each input record, value 1 for each no input record), we calculated an overall value of 0.28 ± 0.27 (median: 0.3; 95% confidence interval: from 0.18 to 0.629). The transformed DRV were entered within the WinBUGS environment after model specification, assuming as success (y = 1) each procedure of input record, and as failure (y = 0) each procedure of no input record. A unequivocal convergence was obtained after 10,000 iterations, and a simulation run was launched for a further 10,000 updates. We obtained a negligible Monte Carlo error and a fine profile in the kernel density plot. This study supported the application of simulation models to databases concerning liver transplantation as a useful strategy to identify a critical state in the data entry process.
|Number of pages||3|
|Publication status||Published - Jul 2007|
ASJC Scopus subject areas