The aim of this retrospective study was to evaluate the survival and associated factors for the longevity of direct posterior restorations and to verify whether the geographic location of public health units could influence the long-term survival of such restorations. Data were extracted from electronic patient files of the Brazilian public oral health services. The sample comprised 2,405 class I and II restorations performed 4 to 24 years ago (mean, 8.9 years) in 351 patients (6.8 teeth/patient) across 12 public health units located in different city regions (42 professionals-55 restorations). The restoration was considered successful if it had not been repaired or replaced at the time of evaluation; failure was defined as replacement of the restoration, the need for endodontic treatment, tooth/restoration fracture or tooth extraction. Data were analyzed using the Kaplan-Meier test for restoration survival and Cox regression to evaluate the factors associated with failure. The majority of the restorations involved the use of amalgam (85%), involved a single face (70%), and were without pulp/dentin capping (85%). The overall survival rate was 95%, and the mean observation time was 8.9 years. The restoration survival was 79% (95% CI: 60.6-89.5) over 24 years, and the mean survival time was 22.2 years (95% CI: 21.9-22.6 years). The annual failure rate up to 24 years was 0.9%. After the adjustment, only the number of restored faces and the geographic location where the restoration was performed remained associated with failure of the restoration. The direct posterior restorations performed at the evaluated public health service units presented high survival rates. The restorations of people with lower access to POHS had lower survival rates. Class I restorations presented higher survival rates than class II restorations with two or more faces, regardless of the restorative material used.