Category Archives: Fall 2016

Patient segmentation and prediction of patient behaviors using the EMR and Machine Learning

In a recent CHOT study, researchers found healthcare organizations can effectively target specific segments of joint replacement patients with readily available information from the electronic medical record (EMR). The goal of this study was to apply unsupervised learning (clustering) methods to a large subset of the patient data to identify unique clusters of patients. By using a new algorithm developed by CHOT researchers, clinicians, can adopt specific protocols during the perioperative phases of Total Joint Arthroplasty by patient cluster, health insurers can isolate high costs and high-risk patients and then incentivize them and their primary care providers to reduce health risks, and patients can learn from their segment what actions to take now to optimize outcomes. Health organizations don’t need to purchase expensive market data to fully understand how patients cluster and respond to health promotions. This innovative algorithm presents an opportunity to efficiently and effectively target segments of the market with health promotions tailored specifically to positively impact health outcomes using existing data in the EMR.

Swensn, E. R., Bastian, N. D., & Nembhard, H. B. (2016). Data analytics in health promotion: Health market segmentation and classification of total joint replacement surgery patients. Expert Systems with Applications60, 118-129.

 

Research Contact:
Eric Swenson
ers187@psu.edu

A new and more accurate model for predicting compliance to evidence-based bundles of practice

CHOT researchers used probability generating functions to rapidly predict compliance and illustrate the improved detection of non-compliance over other methods. This new prediction model for monitoring compliance is practical, fast, and accurate compared to more commonly used types of compliance control charts based on binomial or normal distribution of compliance data, which are shown to produce poor detection properties. CHOT researchers also show that by using this correct method, exact results can be obtained fairly easily, in turn accelerating improvement towards greater compliance to evidence-based practice bundles. Compliance to evidence-based practices, individually and in ‘bundles’, remains an important focus of healthcare quality improvement for many clinical conditions.

Chen, B., Matis, T., & Benneyan, J. (2016). Computing exact bundle compliance control charts via probability generating functions. Health Care Management Science, 103-110.

 

Research Contact:
James Benneyan, PhD
benneyan@coe.neu.edu