Healthcare Analytics

Our mission is to improve healthcare delivery and patient outcomes, by developing, implementing, and disseminating innovations based on data science and operations research. We provide strategic and operational decision support, pioneer Healthcare Systems thinking, and promote positive change.

About us

Healthcare Analytics actively contributes to drive substantial improvements in patient access, flow, quality and efficiency of care. We use multiple methods including advanced analytics (i.e., regression, classification, clustering, text mining, etc.), mathematical optimization (i.e., integer, non-linear, stochastic, etc.) and simulation modeling (i.e., agent based, discrete event, etc.). We work in an engaging environment with clinical partners as well as with operational/administrative stakeholders to ensure our solutions are properly designed, deployed and tested. Healthcare analytics combine professionals from data analytics, operational modeling and applied research.

Data analysis: Creating knowledge from data

Geisinger poses a unique situation as one of the early adopters of electronic health records in the United States. We currently have access to a vast source of information that includes its hospitals, a network of primary care and specialty clinics, and its own health plan. Healthcare Analytics’ data analysts answer operational and research questions by designing, constructing, and validating relevant datasets from numerous available sources. Additionally, our analysts continuously study the diverse data sources within Geisinger to develop standard data sets and draw descriptive statistics that can be used for future studies and analysis.

Operational Modeling: Developing Innovations

Healthcare Analytics’ modelers apply engineering theories, methodologies and tools to develop solutions. The design step brings together clinical partners, stakeholders and end users, ensuring the solutions are effective and easy to use. The most common approaches used are simulation, predictive modeling and optimization. Geisinger hospitals, service lines, departments, and clinics benefit from highly innovative information technology applications that support daily operations and strategic decisions.  

Research: Contributing to Generalizable Knowledge

Healthcare Analytics creates an engine that spurs innovative research, to develop a better understanding of systems engineering concepts and applications in healthcare, so that providers and patients can enjoy their benefits. Healthcare Analytics teams engage in translational research focused upon demonstration, diffusion and dissemination of systems solutions, tools, technologies, models and knowledge that can be generalized and scaled throughout the healthcare delivery system. Our team has been successful in securing external funding from sources such as the Agency for Healthcare Research and Quality (AHRQ), as well as other collaboration-based grants with universities.

Our Commitment to Education

Healthcare Analytics is a catalyst for and leader in healthcare systems engineering education by training current and future engineering and clinical leaders through academic, university, governmental and industry partnerships for research and development. Healthcare Analytics utilizes the entire Geisinger Health System as a research and educational laboratory. Our academic partnerships provide opportunities for undergraduate/graduate experiences, internships, and mentoring. We also perform educational outreach through presentations to courses and student groups, giving working healthcare applications of the concepts they are learning. Healthcare Analytics provides a platform for educational dissemination that contributes to the scaling and generalizing of healthcare solutions.


Featured Projects

Predicting No-shows & Late Cancellations to Improve Clinic Utilization
Patients scheduling but not completing appointments is a common problem in healthcare. This often results in compromised patient care, wasted clinical resources, and limited access for other patients. Most of the current strategies to manage “no-shows” are proven to be ineffective and costly when applied to the entire population, and few efforts have been made to accurately target and intervene on patients likely to no-show. This research aims to construct models using advanced machine learning to predict the risk of patients not completing a given appointment. The risk value is then used to either prevent or mitigate the impact of the risk through preventative and scheduling interventions to improve patient care, provider utilization, and patient access to care.

Inpatient Early Warning Score
This research will focus on developing, validating, and verifying a predictive model to identify non-ICU patients at risk for Cardiac Arrest, ICU transfer, and Death during their inpatient hospitalization. This model can be used to investigate and mitigate these potential events. This Geisinger model will be benchmarked against other real-time inpatient risk algorithm performance.

Strategic Bed Analysis Model (StratBAM)
The ability to accurately measure and assess current and potential health care system capacities is an issue of local and national significance. The Institute of Medicine and the Agency for Healthcare Research and Quality have emphasized the need to apply industrial and systems engineering principles to improving health care quality and patient safety outcomes. To address this need, a decision support tool was developed for planning and budgeting of current and future bed capacity, and evaluating potential process improvement efforts. The Strategic Bed Analysis Model (StratBAM) is a discrete-event simulation model created after a thorough analysis of patient flow and data from Geisinger’s electronic health records. The model is generalizable and can be appropriately scaled for larger and smaller hospital settings.

Advanced Census Analytics
Having a comprehensive view and understanding of total bed demand and supply allows our leadership teams to make a variety of data driven bed decisions. Our specialized census and capacity data marts, visualizations and analyses enable detailed stratification of our patient demographics, units, occupancy, patient and payor mix, bed demand and more. This is used to derive the vital information necessary to construct an analytics question, perform root cause analysis, answer very specific census related questions, eliminate solutions, provide a series of recommendations and create and measure the impact of what-if scenarios in an efficient, accurate manner. Our advanced census analytics are scalable across the organization and have been utilized numerous times for a variety of reasons by our clinical and business leaders while we continue to expand its application at Geisinger.

Predicting and Mitigating Near-Term Inpatient Bed Demand
High occupancy rates increase the probability of adverse events. In the U.S., the inpatient census averages an alarming 60%. At Geisinger we put a great emphasis on patient outcomes, and therefore, inpatient bed planning is a critical aspect which requires better strategies for data-driven decisions. To deliver a timely and highly accurate estimate of the inpatient bed demand, our Team developed a forecasting model based on the application of machine learning algorithms. The new model reduces the mean average error (MAE) error in 40% from MAE= 16.27 (old model) to a MAE=9.8 (New model). Currently, the newly developed forecasting model is being daily used by the Bed Planning Team at the Geisinger Medical Center in central Pennsylvania.

Reducing 72-Hour Post Discharge Adverse Events, Development of REDD Model
Transitional care interventions can be utilized to reduce post-hospital discharge adverse events (AEs). However, no methodology exists to effectively identify high-risk patients of any disease across multiple hospital sites and patient populations for short-term postdischarge AEs. The 3-day REDD (Readmissions, ED Visits, and Deaths) model predicts high-risk patients with fair discriminative power. The discriminative power of the 30-day REDD model is also better than the previously reported models under similar settings. The REDD model has been implemented and is being used to identify and mitigate patients at risk for AEs.