The cross-referencing of data on long-term illnesses (ALD) and the Medical Information Systems Program (PMSI) constitutes the first key focus of the reorientation of the multi-source cancer surveillance system recommended by the pilot study’s findings. The objective of the cross-referencing was to describe, at the national level, the ALD cancer data from the 2006–2008 national cross-insurance system database, the 2004–2008 PMSI hospital stays for cancer, as well as the data resulting from the cross-referencing of these two sources over the 2006–2008 period. The aim was also to develop a method for constructing surveillance indicators based on the cross-referenced data. Eight cancer sites (breast, lung, colorectal, cervix and uterine body, thyroid, central nervous system (CNS), and kidney) were studied. Hospital stays for cancer during the study year were selected using ICD-10 codes for the primary diagnosis, linked with identical stays from the PMSI database for previous years. Patients with long-term illness (ALD) due to cancer were selected using the ICD-10 code and the earliest ALD start date. The selected cases were cross-referenced using a common anonymous identifier. The indicators measuring their levels were compared with national incidence estimates. The relative difference between the indicator derived from the cross-referencing and the estimated incidence in 2008 was: - very small for lung (-1%) and breast (+6%); - significant for thyroid (+17%), colorectal (+22%), kidney (+28%), and uterine body (+33%); - very large for the uterine body (+33%), the cervix (+49%), and the CNS (+61%). The large relative discrepancy observed for certain cancers is likely due to a high number of "false positives" among the unpaired cases. The PMSI and ALD systems were not designed for epidemiological purposes but rather for medico-economic ones. However, their national coverage, the comprehensiveness of the data, and the common patient identification number constitute major assets for cancer surveillance. (R.A.)