This study proposes a hierarchical and explainable diagnostic approach to accurately identify bacterial vaginosis from patient health, pathogen presence, and vaginal microbiota composition. Three explainable models are built from highly accurate black-box models, providing healthcare professionals with technological support to improve decision-making in the follow-up and treatment of patients with suspected bacterial vaginosis. The aim is to reduce diagnostic errors and optimize the use of medical resources in clinical settings. Experimental studies suggest that decision trees as explainers for black-box model predictions outperform the results obtained when directly using a white-box model on the same data. The proposed approach allows for improved diagnostic reliability and helps in the interpretation of results by physicians, increasing the confidence of using artificial intelligence tools in clinical practice.