Study design

This retrospective study analyzed data extracted from the standard French national hospital discharge database, known as the PMSI (Programme de Médicalisation des Systèmes d’Information), [4] over the 2018–2019 period (Declaration number 2,203,389 v.0).

Data selection

In France, the PMSI database records all procedures performed during all hospitalizations, including daycare or longer, in private or public institutes [4]. Since July 2017, MT has become a well-defined, specific procedure coded as EAJF341 under the common classification of medical acts (Classification commune des actes médicaux, CCAM) and is thus coded as such in the PMSI database. To study two complete years, all data covering 2018 and 2019 were extracted. Data on patients aged 20 or over with “stroke” as the main diagnosis (International Classification of Diseases code 10 : I63.0) associated or not with MT were extracted for inclusion in the study. The code specific to thrombectomy was then checked. Both initial and recurrent strokes – apart from early recurrences, i.e. recurrences occurring during the ongoing hospital treatment phase for a first incidence - were included.

Endpoints and assessments

The primary endpoint was the number of MTs performed across the territory, standardized per age, sex and percentage of strokes. Secondary endpoints were the mortality rate at 1 year after thrombectomy, and associated risk factors.

Depending on the number of MTs performed in 2018 and 2019, and according to the presumed experience of the operators, stroke centers were classified as follows [5] :

very small MT centers with very few thrombectomies performed (1 to 9 MTs/year): no experience.

small MT centers with few thrombectomies performed (10 to 49 MTs/year): low experience.

occasional MT centers with 50 to 99 MTs/year (> 1/week): regular experience.

regular MT centers with 100 to 199 MTs/year, (2 to 4/week): good experience.

high-volume MT centers with 200 to 399 MTs/year (more than 4/week): very good experience.

very high-volume centers with more than 399 MTs/year (> 1/day): experts.

Endpoints concerning mortality were death rates during hospital stay and at 1 year. Data concerning comorbidities including chronic heart, respiratory or renal failure, cancer, diabetes mellitus and cardiovascular pathologies were also extracted from the PSMI to assess prognostic factors. The study on mortality only focused on patients who had undergone thrombectomy.

Statistical analysis

Statistical analyses were carried out using R software (version 4.2.2).

Standardized prevalence was calculated with the number of MTs for 100 000 inhabitants per year and per department, with standardization according to sex and age. The reference population was defined with the number of inhabitants in France over the period.

A generalized linear mixed model was used to estimate the one-year probability of death for patients undergoing thrombectomy (death during hospital stay or subsequent visits). To account for potential differences in mortality between centers, centers were treated as random effects. All variables characterizing patients and their hospital stays: length of stay, age, gender, Charlson comorbidity index, the 17 components of the Charlson index, center size (in terms of the number of thrombectomies performed), month, and year were treated as fixed effects. The linearity of quantitative variables were assessed and transformed to achieve a linear effect. The natural logarithm of the length of stay and the square root of the number of thrombectomies were thus taken. The final multivariate model included the variables of interest in the study, as well any variables that emerged as being significant through the likelihood-ratio test.