RISK-ADJUSTED MONITORING, COST-EFFECTIVENESS, AND META-ANALYSIS FOR DATA-DRIVEN HEALTHCARE FOR TOTAL HIP ARTHROPLASTY
Open Access
- Author:
- Yu, Yifeng
- Graduate Program:
- Industrial Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- February 25, 2017
- Committee Members:
- Harriet Nembhard, Dissertation Advisor/Co-Advisor
Harriet Nembhard, Committee Chair/Co-Chair
Conrad Tucker, Committee Member
Ling Rothrock, Committee Member
Naleef Fareed, Outside Member
Andrea Yevchak-Sillner, Special Member - Keywords:
- total hip arthroplasty
risk-adjusted monitoring
meta-analysis
machine learning
statistical process control
cost-effectiveness
Markov simulation model - Abstract:
- Total hip arthroplasty (THA) is the most effective surgical intervention for patients with severe hip arthritis or hip injury. It has been shown to significantly improve patients’ quality of life with respect to physical function, pain relief, and overall health. In recent years, unplanned readmission after THA has become an increasingly serious problem in the United States. Unplanned readmission significantly reduces the quality of life of THA patients. Moreover, in fiscal year 2015, the Centers for Medicare and Medicaid Services (CMS) started to penalize hospitals for high 30-day readmission rates after THA surgery. Relevant studies indicated that readmission after THA surgery is deemed a healthcare quality indicator and that a considerable portion of 30-day unplanned readmission after THA can be effectively prevented. Therefore, it is necessary to mitigate the potential problems and improve THA care quality in order to reduce unplanned readmission. Moreover, the demand for THA surgery is expected to increase from 326,100 in 2010 to 572,000 in 2030 in the United States. However, the available capacity of THA may be insufficient for such a massively increasing demand. In this landscape, effectively using THA surgery to optimize patient outcome becomes an important topic. With the advancement in healthcare information technology, healthcare data has become exponentially available. Data science and engineering is needed in the healthcare ecosystem to support medical decision making, facilitate quality improvement initiatives, and improve patient outcomes. However, data science and engineering has not been adequately used in improving THA patient outcomes. The objective of this doctoral dissertation is threefold: 1) to advance data-driven healthcare for THA, 2) to provide insights into informed medical decision making, and 3) to support and guide efforts in THA care quality improvement. Since 30-day unplanned readmission has become an increasingly serious problem and hospitals now bear the risk of CMS penalties, it is very important for hospitals to understand their THA surgical performance in real time. We obtained identified THA patient-level data records from an academic medical center in central Pennsylvania, United States. We proposed using machine learning algorithms to conduct patient risk stratification, and combining them with statistical process control to perform surgical outcome evaluation. The results indicate that random forest outperforms the most commonly-used risk-adjustment method, logistic regression, in identifying high-risk patients. Therefore, the control chart based on random forest provides more convincing results on surgical outcome evaluation. With our proposed risk-adjusted monitoring framework, the medical team can better target interventions on future high-risk patients, and diagnose potential care quality problems in a timely manner. Due to the possible shortage of THA surgery in the next 15 years, it is important to take cost-effectiveness into account when determining the appropriate treatment option for each patient. We proposed a method for clustering hip arthritis patients and analyzing the cost-effectiveness of THA surgery for each patient subgroup. The results indicate that THA surgery is more cost-effective for relatively young patients with few co-morbidities, while it is not cost-effective for the oldest patient group in this study. With our proposed method, coupled with the characteristics of each patient subgroup, the medical team can better determine the optimal treatment option for each hip arthritis patient in order to improve patient outcomes. Finally, when implementing medical or surgical interventions on THA patients, convincing evidence must show that a certain intervention really benefits the patients. The problem, however, is that relevant studies show mixed results on the effectiveness of a certain intervention. To address this issue, we used meta-analysis to deliver a valid and comprehensive understanding of the effectiveness of an intervention and we analyzed pre-operative exercise as an example. The results indicate that while pre-operative exercise provides moderate benefits with respect to pain relief, physical function, and activity of daily living before surgery, these benefits diminish and become insignificant post-operatively. Therefore, our meta-analysis provides important insights into what medical or surgical interventions have proven effectiveness in improving THA care quality and patient outcomes, based on the mixed results of existing literature. In conclusion, this data-driven approach to THA advances the translation of data science and engineering in healthcare, by using the increasingly available healthcare data to deliver insightful solutions. This approach can be used to facilitate efforts in THA care quality improvement, with the purpose of managing care quality and improving THA patient outcomes.