AFFILIAZIONE
new delta
AUTORE PRINCIPALE
Dott.ssa Bruno Martina
Registrazione obbligatoria. Una valutazione consentita
login avvenuto
GRUPPO DI LAVORO
Dott.ssa Bruno Martina new delta
Dott.ssa Moncado Francesca università degli studi di palermo
Dott.ssa Ferrito Maria Federica università degli studi di palermo
PhD La Rosa Martina
PhD Sicari Mirea università degli studi di messina
PhD Calanni Pileri Michela humanitas university
Dott.ssa Moncado Francesca università degli studi di palermo
Dott.ssa Ferrito Maria Federica università degli studi di palermo
PhD La Rosa Martina
PhD Sicari Mirea università degli studi di messina
PhD Calanni Pileri Michela humanitas university
AREA TEMATICA
Applicazioni di intelligenza artificiale in sanità
ABSTRACT
Continuous monitoring of vital signs combined with early risk prediction models has shown significant potential in improving patient outcomes and optimizing healthcare resources (Zhao et al., 2020; Johnson et al., 2018). PredictCare is a web-based platform designed to simulate an advanced multiparametric monitoring system, integrating real-time data acquisition and AI-driven risk stratification. The primary objective is to develop a software-as-a-medical-device (SaMD) prototype capable of real-time vital signs tracking, automated clinical risk assessment, and enhanced patient-provider communication via an AI-powered interface. PredictCare simulates the acquisition of core clinical parameters (systolic and diastolic blood pressure, heart rate, SpO₂, body temperature, BMI, and stress indices) using synthetic patient data.
A two-stage XGBoost classification model was developed, achieving a recall of 98% for critical patients (Stage 3), in line with best practices for high-sensitivity medical triage systems (Rajkomar et al., 2019).
The platform also incorporates an interactive dashboard, automated PDF report generation, and HealthGuard AI — a conversational agent based on OpenAI’s GPT-4 architecture — to facilitate patient interaction and symptom documentation
PredictCare demonstrated robust performance: Overall model accuracy of 95%; Precision (macro average): 88%; Recall (macro average): 89%; F1-Score (macro average): 88%.
The architecture prioritizes critical patient detection while minimizing false negatives, supporting early clinical intervention. The system allows dynamic tracking of clinical variables, personalized alerts, and health trend visualization, meeting the need for scalable, AI-enhanced remote patient monitoring solutions.
PredictCare represents an innovative step towards data-driven, AI-supported healthcare, offering a modular framework for future integration with wearable devices and electronic health records (EHRs). Further development aims to pursue regulatory compliance and clinical validation.
Although currently not certified as a medical device, PredictCare highlights the transformative role of AI in precision medicine and proactive patient care.
A two-stage XGBoost classification model was developed, achieving a recall of 98% for critical patients (Stage 3), in line with best practices for high-sensitivity medical triage systems (Rajkomar et al., 2019).
The platform also incorporates an interactive dashboard, automated PDF report generation, and HealthGuard AI — a conversational agent based on OpenAI’s GPT-4 architecture — to facilitate patient interaction and symptom documentation
PredictCare demonstrated robust performance: Overall model accuracy of 95%; Precision (macro average): 88%; Recall (macro average): 89%; F1-Score (macro average): 88%.
The architecture prioritizes critical patient detection while minimizing false negatives, supporting early clinical intervention. The system allows dynamic tracking of clinical variables, personalized alerts, and health trend visualization, meeting the need for scalable, AI-enhanced remote patient monitoring solutions.
PredictCare represents an innovative step towards data-driven, AI-supported healthcare, offering a modular framework for future integration with wearable devices and electronic health records (EHRs). Further development aims to pursue regulatory compliance and clinical validation.
Although currently not certified as a medical device, PredictCare highlights the transformative role of AI in precision medicine and proactive patient care.
Zhao, M. et al. 2020. DOI: 10.3389/fmed.2020.637434
Johnson et al., 2018. DOI: 10.1109/JPROC.2015.2501978
Rajkomar et al., 2019. DOI: 10.1056/NEJMra1814259
