STAGE 7 SPOTLIGHT RUMC: Improving the Clinical Workflow

HIMSS Analytics

The HIMSS Analytics Adoption Model for Analytics Maturity (AMAM) helps healthcare providers around the world focus their analytics efforts and investments by driving clinical, operational and financial analytics efforts to align with and support their organizational strategy. The model is flexible enough to accommodate different types of organizations, cultures and approaches to maturity. It is a staged roadmap designed to help providers successfully focus and align analytics efforts through a vendor neutral, capability oriented approach.

Rush University Medical Center (RUMC) used the AMAM to achieve impressive new and increased efficiencies in the overall care process, achieving the final stage of maturity (Stage 7) in December 2018. Here, we follow the RUMC journey to implementing the model for the hospital-wide system.

The Challenge

When it comes to emergency room performance, one of the most significant indicators is the number of visitors that leave the Emergency Department (ED) without being seen by practitioner. The number of unseen ED visitors is denoted by Leave Without Being Seen (LWBS) and is impacted by several factors, including the duration of inpatient stay without transfer, the cycling of patients throughout the hospital and overall emergency room efficiency.

Wait time in the ED reflects how well the resources are used within and outside of the department. The more visitors leave the ED, the higher the possibility that resources are not being used efficiently and effectively throughout the entire hospital system. Additional consequences could be “suboptimal care and medicolegal risk,” according to Paul Casey MD, Senior Patient Safety Officer and Associate CMO at Rush.

In addition, studies show that unattended visitors tend to return within two days, with worsened symptoms (which in turn impacts the financial burden of the hospital). To improve hospital system efficiency and patient satisfaction, RUMC set a goal for LWBS for FY’ 19 of 1.4% of ED visitors, developed based on a baseline rate of about 4.4%.

Implementation

The focus of the LWBS reduction campaign centered around increasing efficiency in the overall care processes. The project of integrating predictive modeling into the clinical workflow was headed by ED practitioners (of various specialties), executives and data specialists. The team identified visitors most likely to leave the ED, customizing the workflows toward quick intervention.

The team utilized predictive and prescriptive analytics to develop predictive models, using a “sequential process of data set extraction, machine learning model training, model refinement through user validation, development of a deployment framework in the Epic EHR, and production deployment,” according to the original RUMC Case Study. The initial model version encompassed factors including medication, billing and encounter history, along with laboratory tests and other patient health data, integrating 136 different features to address all areas. The 136 features involved in the original model were reduced to all-encompassing 8, resulting in a Recall of 86%.

All the real-time data was leveraged by Epic’s Cognitive Computing Platform (ECCP), which implemented the predictive models into the administrative and clinical workflow. The resulting models predicted the probability of LWBS of each ED visitor. Workflows were adjusted based on the outputs of the models to carry out preventative methods when needed.

Results

The original LWBS rate left about 3,375 ED patients without treatment annually, with RUMC losing about $1,000 per patient. The integration of LWBS predictive models into the clinical workflow resulted in significant ROI for RUMC. After the integration campaign allowing practitioners to identify visitors who are likely to leave, the LWBS in the ED fell from 3.83% to 1.25% — extremely close to the aggressive 3% reduction goal at the beginning of the campaign. The hospital gained about $1.5 million per year as a result of the decreased LWBS rate.

The Takeaway

The development of the LWBS predictive model allowed RUMC to strengthen the hospital-wide collaboration culture, address technical challenges within the hospital system and implement a technology with a high ROI. The campaign resulted in increased emergency room efficiency and the reinforcement of the hospital's overall mission — to improve health by leveraging the people, programs, reach and community of the overall hospital system.

case study