Introduction. Group B Streptococcus (GBS) remains the leading cause of bacterial neonatal infections worldwide, despite the spread of recommendations on vaginal screening and antibiotic prophylaxis. Hypothesis/Gap Statement. There is a need to evaluate the potential changes in GBS epidemiology over time following the introduction of such guidelines. Aim. Our aim was to perform a descriptive analysis of the epidemiological characteristics of GBS by conducting a long-term surveillance of strains isolated between 2000 and 2018, using molecular typing methods. Methodology. A total of 121 invasive strains, responsible for maternal infections (20 strains), fetal infections (8 strains) and neonatal infections (93 strains), were included in the study, representing all the invasive isolates during the period; in addition, 384 colonization strains isolated from vaginal or newborn samples were randomly selected. The 505 strains were characterized by capsular polysaccharide (CPS) type multiplex PCR assay and the clonal complex (CC) was assigned using a single nucleotide polymorphism PCR assay. Antibiotic susceptibility was also determined. Results. CPS types III (32.1 % of the strains), Ia (24.6 %) and V (19 %) were the most prevalent. The five main CCs observed were CC1 (26.3 % of the strains), CC17 (22.2 %), CC19 (16.2 %), CC23 (15.8 %) and CC10 (13.9 %). Neonatal invasive GBS diseases were predominantly due to CC17 isolates (46.3 % of the strains), which mainly express CPS type III (87.5 %), with a very high prevalence in late-onset diseases (76.2 %). Conclusion. Between 2000 and 2018, we observed a decrease in the proportion of CC1 strains, which mainly express CPS type V, and an increase in the proportion of CC23 strains, mainly expressing CPS type Ia. Conversely, there was no significant change in the proportion of strains resistant to macrolides, lincosamides or tetracyclines. The two molecular techniques used in our study provide almost as much information as classical serotyping and multilocus sequence typing, but are quicker, easy to perform, and avoid long sequencing and analysis steps.