Optimizing hospital decision making through integrating the Systems Engineering Initiative for Patient Safety (SEIPS) model with data-driven methods

Abstract

Patient safety remains a critical priority in healthcare, yet hospitals continue to face challenges in improving patient outcomes, especially for patients enrolled in clinical trial research. Clinical trials can bring additional risks and financial burdens to patients, and integrating all relevant information into a single management system remains a challenge. Patient outcome is a key indicator of hospital performance in both patient safeties. Traditional patient safety strategies rely on statistical analysis but fail to proactively mitigate risks using real-time patient status. Integrating data-driven approaches and artificial intelligence (AI) can transform hospital decision-making through real-time risk assessment and early intervention. However, limited research has explored how predictive modeling can be effectively incorporated into hospital system to enhance patient safety. This dissertation integrates the Systems Engineering Initiative for Patient Safety (SEIPS) framework with AI-driven predictive modeling to streamline the development and implementation of prediction models in inpatient care. The study employs logistic regression and machine learning techniques to predict patient outcomes, focusing on mortality, prolonged hospital stays, and total charges. Three objectives guide this research: (1) identifying key risk factors for pediatric patient mortality in clinical trials using a national inpatient database, (2) developing a logistic regression model for adult in-hospital mortality prediction in a rural hospital setting, and (3) designing a dynamic, real-time patient outcome prediction model using machine learning methods. The results reveal significant associations between patient demographics, diagnosis, hospital resources, and outcomes. Objective 1 identified key mortality predictors using a national database, including the significant link between length of stay and mortality. Objective 2 highlighted critical predictors of clinical trial patient mortality in rural hospitals. Lastly, Objective 3 demonstrated an AI-based dynamic model for concurrent patient outcome prediction outperformed static models. By integrating predictive analytics into an adapted SEIPS framework, a structure used to develop and implement to developing and implementing patient outcome prediction models, improving patient safety and hospital efficiency. Future work should refine AI models for greater transparency and clinical implementation while addressing ethical and regulatory challenges in AI-driven healthcare solutions.