International Journal of Drug Delivery Technology
Volume 16, Issue 1s

Predictive and Prescriptive Analytics for Medical Treatment Assignment: A Data-Driven System for Pain Management Optimization at ALIVIUM SAS

Frantz José Colimon Gómez 1, Luis Fernando Delgado Morales 2, David Alberto García Arango 3, Leidy Catalina Acosta Agudelo 4, Mauricio Stiven Echeverri Gutierrez5, Camilo Andrés Echeverri Gutiérrez6

1ALIVIUM S.A.S. ORCID: https://orcid.org/0009-0007-8001-7636 email: financiero@ceadalivium.com
2ALIVIUM S.A.S. ORCID: https://orcid.org/0009-0000-6130-3089 email: gerencia@ceadalivium.com,
3Administrative Management Consultants S.A.S ORCID: https://orcid.org/0000-0002-0031-4275, email: investigacion@amyc.com.co
4Administrative Management Consultants S.A.S ORCID: https://orcid.org/0000-0003-1538-1881, email: comercial@amyc.com.co
5Administrative Management Consultants S.A.S ORCID: https://orcid.org/0000-0003-0730-0121 email: gerencia@amyc.com.co
6Administrative Management Consultants S.A.S ORCID: https://orcid.org/0000-0003-0667-0913, email: presidencia@amyc.com.co


ABSTRACT

Ineffective treatment assignment in pain management leads to unnecessary procedures, increased costs, and suboptimal patient outcomes. Traditional approaches often fail to consider non-medical factors that significantly influence treatment efficacy. This study presents the design, development, and evaluation of an integrated predictive and prescriptive analytics system for optimizing medical treatment assignment at ALIVIUM SAS, a specialized pain management institution in Colombia. We developed a two-service interoperable system comprising a software layer with modular monolith architecture and a comprehensive data system integrating Data Lake, Data Warehouse, and machine learning microservices. The system employs multiple ML algorithms (Logistic Regression, Random Forest, XGBoost, Neural Networks, and Ensemble methods) trained on 3,850 patient records augmented through linear interpolation, Gaussian noise injection, and fixed algorithm techniques. Development followed a hybrid Waterfall-to-Agile methodology incorporating Test-Driven Development (TDD), DevSecOps, MLOps, and robust data governance practices. The Ensemble model achieved superior performance with 92.1% accuracy, 91.5% precision, 92.3% recall, 91.9% F1-score, 91.8% specificity, and 0.935 AUC-ROC. System performance metrics demonstrated 245 ms average API response time, support for 500+ concurrent users, 98.5% data completeness, and 99.9% authentication success rate. Feature importance analysis revealed pain intensity score (18.5%), treatment history (16.2%), and comorbidities (14.8%) as primary predictors, while non-medical factors (social, labor, cultural) contributed 31.5% collectively. The implemented system demonstrates the feasibility and effectiveness of integrating predictive and prescriptive analytics into clinical workflows for pain management. The architecture supports real-time decision support while maintaining data security, privacy, and explainability. This work contributes methodological insights on data augmentation strategies, MLOps implementation in healthcare, and the integration of medical and non-medical factors for treatment optimization.

Keywords: Predictive analytics, prescriptive analytics, machine learning, pain management, clinical decision support systems, healthcare informatics, treatment optimization, data governance, MLOps

How to cite this article:Gómez FJC, Morales LFD, Arango DAG, Agudelo LCA, Gutierrez MSE, Gutiérrez CAE., Predictive and Prescriptive Analytics for Medical Treatment Assignment: A Data-Driven System for Pain Management Optimization at ALIVIUM SAS .Int J Drug Deliv Technol. 2026;16(1s): 964-981; DOI: 10.25258/ijddt.16. 964-981