Abstract:
|
Using administrative data, such as tax records or health administrative data are common in demographic or economic area to construct longitudinal individual pathways. Administrative data are only observation of using services. Hence, algorithms are used to encode administrative data into individuals pathways. Each admistrative pathway is divided into subsequences administrative records to construct individual states trajectories. To improve this automatical encoding, target trajectories are also considered for the same admistrative pathways. Each sequence cutting is considered as a partition. Rand index is used to mesure the superimposed of subsequences between algorithms and target one . Due to the the case when algotihm partition is largest than the gold standard target, an asymetric mesure is used. An empiral test is developped in the framework of the comparison of two pathways to compare the value of asymetric rand index value An application to encode patient pathways from claims data is presented.
|