Patient Segmentation

Background

EdjAnalytics and a regional hospital system operating in Kentucky and Indiana partnered to develop solutions that illuminate patterns in patient behaviors. In the face of increasing competition for loyalty within the healthcare industry, the hospital system’s vision is to become more consumer-centric and deliver patients more relevant touchpoints based on their unique needs. Edj was engaged to provide a framework for how the hospital system can better serve and communicate with their patients.

Approach

To support the segmentation development, the hospital system provided data on 1.9 million patients, including Demographics such as Age, Race and Gender; Clinical History such as Diagnoses, Encounters and Medications; and Engagement Metrics such as Online Portal Usage and Patient Satisfaction. The Edj team ⎯ made up of a product manager, data scientist, data analyst and database administrator ⎯ selected a clustering methodology to build a patient segmentation unique to this hospital system. The methodology places patients into groups whereby those in the same group exhibit similar characteristics to each other and differ from the characteristics of those in other groups.

Results

Topological Embedding: Mapping Segments onto a 2D Space

The elements Edj found most useful in discerning patient segments from one another were Encounters and Diagnosis History, revealing how often patients visit and the severity of the conditions being treated. The segmentation organized patients across a continuum from under-engaged and generally healthy to patients with moderate to severe health risk. Each segment included a balance of demographic groups, indicating that factors like age, race and gender do not dictate service utilization within the client’s healthcare system.

Patients with increased health risk, such as those with multiple advanced chronic illnesses visit much more frequently (32 visits per year on average) compared to those in the healthy group who visit only 3 times per year on average. An additional group composed of healthy maternity patients was added as a specific population of interest for the marketing team at the hospital system given they deliver the largest number of newborns in the region in which they operate.

Full Segmentation Overview

Next Steps

Edj’s hospital system client is now able to use these segments to develop marketing messages and campaigns relevant to each group as well as filter existing reports or dashboards developed by the financial and strategic reporting teams to understand where differences exist by segment. By executing primary research with each group, the marketing team will gain increased insight into the motivations that drive their patient behaviors and potential new services the hospital might offer to better address the needs of each segment.

The two companies are embarking on a second project involving more sophisticated, predictive analytics techniques. The goal of the second project is to identify patients who are likely to need hip replacement, knee replacement and bariatric surgeries. Identifying specific patients who will require these surgeries in the near future will enable the hospital marketing team to target them with educational materials about the benefits of choosing their system as their care provider for those procedures. In the long run, predictive analytics will yield an increased level of personalization, patient conversions and increased market share for the regional healthcare system.