Background: Hospital readmission rates are a critical measure of healthcare quality and patient outcomes. High readmission rates can indicate poor patient management and lead to increased healthcare costs. Innovations in artificial Intelligence (AI) have the potential to improve patient care through clinical decision support tools, generative clinical care pathways, and predictive analytics creating the potential for more personalized, cost-effective healthcare delivery.
Methods: A systematic review of 37 articles from January 2013 to January 2024 was conducted using PubMed. The search focused on studies exploring the application of Machine Learning/AI in healthcare in predicting/reducing hospital readmission rates, including the ethical implications of AI. Key articles were selected to illustrate the diverse applications and outcomes of AI in reducing readmissions.
Results: The systematic review revealed that ChatGPT and other AI tools improved patient communication and education, enhancing adherence to treatment plans and reducing potential readmissions. In community-based primary healthcare, data from 35 studies indicated that AI significantly aids in early diagnosis and continuous disease management, crucial for preventing readmissions. In anesthesiology, a review of 15 applications of AI and telemedicine showed a 20% reduction in post-operative complications through remote monitoring and timely interventions. Improved surgical documentation using GPT-4 led to a 15% reduction in documentation errors, enhancing post-operative care and decreasing readmissions. AI’s 85% accuracy in forensic data analysis was noted, which could be translated to medical contexts for accurate patient outcomes, thus reducing readmissions. Addressing biases in AI systems, as highlighted in 25 studies, is critical for equitable healthcare and reducing readmissions. Comparisons of 30 AI models to human clinicians demonstrated similar accuracy levels but faster processing times, enhancing diagnostic accuracy and patient management, potentially reducing readmissions. AI-driven advancements in COVID-19 diagnostics showed a 25-30% improvement in diagnostic accuracy, which contributed to reduced complications and readmissions. Literature also revealed the use of machine learning models to predict the readmission of patients with COPD exacerbations and Heart failure along with prediction of sepsis.
Conclusions: Our review of the literature shows that the integration of AI into healthcare has shown promising potential in reducing hospital readmission rates by optimizing patient management and predicting health outcomes. Machine learning algorithms have been effective in predicting patient length of stay and identifying patients at risk for early readmission. These predictive capabilities allow healthcare providers to implement tailored interventions, streamline discharge processes, personalize post-discharge care plans, and ensure appropriate follow-up care to reduce readmissions. In conclusion, the strategic deployment of AI in hospital settings not only enhances the precision of patient care but also offers a sustainable approach to reducing readmission rates through predictive analytics and personalized care planning.