Performance and Security in AI-Driven mHealth: A Comparative Evaluation of PostgreSQL, MongoDB, and Firebase for Endometriosis Management

Mobile health (mHealth) applications transform healthcare for chronic conditions like endometriosis. This paper compares Firebase Firestore, MongoDB, and PostgreSQL for “PainCare”, an mHealth application developed for endometriosis symptom tracking and AI-driven predictive analysis. We evaluated these systems on backend performance (using ApacheJMeter with up to 100 users), application security (MobSF static, Burp Suite dynamic), and Random Forest AI model integration (trained on a $\mathbf{1 0, 0 0 0}$-instance Kaggle dataset) for endometriosis likelihood prediction. Our findings show PostgreSQL exhibited superior performance and stability under load $(28-32 ~\text{ms}$ avg. response, 0 % error). Firebase enabled rapid development but faced Firestore performance testing limitations due to SSL errors, alongside a critical vulnerability from an exposed API key. All implementations received “Medium Risk” MobSF security scores (Firebase: 50/100, MongoDB: 44/100, PostgreSQL: 48/100), highlighting common weaknesses like debug configurations. MongoDB offered flexibility for the AI integration, where the Random Forest model achieved approximately 63.27 % accuracy. With 63 % accuracy, the AI model cannot be used for clinical diagnosis, but can serve as an awareness or preliminary triage tool, helping direct patients to medical consultation. Future improvements are needed to strengthen its reliability and applicability. This study provides empirical evidence and practical considerations for DBMS selection in AI-enabled mHealth, highlighting the essential relationship between performance, security, AI feasibility, and robust application-level security practices while supporting global smart connectivity for decentralized healthcare systems.