Pragmatic study comparing historical data from usual collection tools to medical administrative databases (local / national PMSI and CRF / SNIIRAM). There are 3 phases to the data collection:

1. Collection of hospitalization data from the local PMSI and outpatient data from CRF
2. Collection of hospitalization data in the national PMSI with a probabilistic matching between local and national PMSI data
3. Collection of outpatient consumption data from SNIIRAM after matching of data from the national PMSI to the SNIIRAM.

In studies assessing therapeutic interventions, an economic evaluation requires the collection of patients' healthcare consumption. Two methods are primarily used in accordance with the recommendations of the National Authority for Health (HAS): case report forms (CRF) for ambulatory care and the local hospital discharge databases (local PMSI) - extracted from participating centers - for hospital care. However, these conventional methods have limits. The first is a declarative and mostly retrospective collection, often misinformed and unverifiable. The second provides high quality data but is far from exhaustive as it only concerns hospitalizations that occurred in the participating center, not those that occurred in other hospitals.

In France, medical administrative databases SNIIRAM (all care reimbursed by the statutory health insurance) and national PMSI (all hospital care) are the "gold standards" for collecting consumer personal care. However, the administrative difficulties associated with accessing these databases prevent researchers from extracting routine care consumption data from these databases for health economic studies, despite the fact that the unreliability of the conventional methods has been objectified in a small number of studies in the literature. These studies are few, have been conducted abroad and seek to compare declarative collections to non-exhaustive medical and administrative data. Moreover, they compare only the number of events collected and not the implication for economic evaluations using those numbers, limiting the applicability of their results to economic assessments.

The main objective of this study is to evaluate the performance of usual collection tools compared to the gold standards (medical and administrative databases) for estimating consumption of inpatient and outpatient care in health economic evaluations. The secondary objectives are to evaluate the impact of the potential underestimation of healthcare consumption on the estimated costs and results of economic assessments by studying the variation of the calculated incremental cost-effectiveness ratio (ICER) and the conclusion of evaluation.

The COMERE population comes from studies in which an economic evaluation for an innovative therapeutic strategy was conducted by the principal investigators's research unity. These studies include at the minimum a collection of hospital care consumption. This data collection must be over by the time for the start of the project. The procedure code and studied pathology alone allow the investigators to identify patients in the national PMSI with a probability greater than 0.9. When adding other information available in the local PMSI, this probability reaches 0.99. The principal investigators of the studies agreed on the use of data from their studies in the context of this research project. Overall, among the studies carried out in our unit, 12 projects funded by the Ministry of Health's program supporting funds for expensive innovative technology (STIC) and 1 national registry matched the study's criteria and were retained. Of these 13 studies: 13 have collected data on hospital care consumption and 8 collected data on ambulatory care consumption. In total, these studies included 87 separate centers and 6,928 patients of which 2,661 have an ambulatory care collection. Their data are completely anonymous.

Inclusion Criteria:

Among patients of the source population, patients were included if:

• Their initial hospital stay (the one during which the innovative treatment strategy was implemented) was found in the local PMSI
• A probabilistic matching with the national PMSI is possible with a probability equal to 1 (in order to compare the data from the usual collections to the gold standard methods).

Exclusion Criteria:

• We excluded patients whose initial hospital stay found in the national PMSI was not linkable with other admissions due to a coding error.

Patient with an initial hospital stay during which the innovative treatment strategy was implemented