A Smart and Connected Healthcare Delivery Process PDF Download

Author: Sujee Lee
Publisher:
ISBN:
Category :
Languages : en
Pages : 0

Get Book

Book Description
Healthcare delivery is facing a paradigm change to embrace rapid development in information technology, data analytics, artificial intelligence, as well as numerous medical devices and treatments to achieve smart and interconnected care. In a smart and interconnected healthcare system, integration of data analytics, system modeling, optimal decision-making, and care intervention is necessary and important. Based on the collected data, including the patient's demographic information, disease history, physical exam, and diagnostic test, the smart and patient-specific intervention decision will be formed, and proper care practice will be delivered. Through this, all activities of prevention, diagnosis, treatment, clinical visits, and home care are all connected together. This dissertation is dedicated to providing analytical frameworks for such a smart and connected healthcare delivery process to address issues related to classification, prediction, intervention, and care service by integrating machine learning, optimization, and system modeling techniques. Specifically, (1) through data collection and preprocessing, predictive models are developed to stratify patients, determine the patient's status, or predict risks by means of machine learning algorithms. (2) Based on patient identification, through modeling the post-discharge care process, intervention plans and policies are evaluated and optimal decisions are proposed. (3) Finally, to implement the intervention and treatment plan, care delivery policies are studied to improve care quality. Through these steps, an integrated and comprehensive framework can be established to connect data analytics, intervention planning, and care services in a closed loop. In order to show the significance and applicability of such frameworks for the smart and connected healthcare system, this dissertation introduces analytical frameworks applied on readmission risk management for COPD (Chronic Obstructive Pulmonary Disease) patients and opioid prescription optimization for TJR (Total Joint Replacement) patients. In order to provide models of continuous care delivery, a workflow policy development study for primary care physicians is also introduced. Specifically, for reducing COPD readmissions, two different sub-frameworks integrating machine learning models and operations research methods are developed. The first sub-framework classifies COPD patients into the high or low risk of readmissions, then based on the risk group, an intervention resource allocation is determined through linear programming and graphical analysis. In the second sub-framework, a training procedure for a causal Bayesian network is proposed and the resulting causal network describing relationships between factors and readmission is integrated into a Markov decision process to provide a dynamic intervention planning. For opioid prescription optimization, a novel approach for semi-supervised learning is proposed and the resulting classification model predicts patients' expected opioid consumption levels. Then, a stochastic program is introduced to decide how many opioids should be prescribed to each class of patients to reduce opioid leftovers and thereby curtail the opioid crisis. Finally, as primary care is in charge of continuous care delivery regardless of patients' underlying diseases and conditions, workflow models for primary care physicians are developed by utilizing stochastic process modeling techniques. In summary, the work developed in this dissertation provides novel frameworks enabling smart and interconnected care to treat patients in need and resolve issues in the U.S. healthcare system.