Personalized healthcare with predictive analytics

Case study

Personalized healthcare with predictive analytics

In the realm of employee health care, much progress has been made, but the path to clear healthcare costs is still a long road ahead with many hurdles along the way. Large employers offering health care benefits to their employees face difficulty in creating awareness and connecting employees with relevant benefits and services before they make healthcare decisions. Critical factors affecting healthcare costs include:

  • Low employee engagement – 40% of employees were unable to comprehend their health insurance benefits.
  • Underutilized employer health benefits – only 12% of employees were fully proficient with their health insurance benefits.

This results in employees making uninformed decisions, which incurs higher healthcare costs.

Our client offers a B2B healthcare information platform that optimizes annual health insurance costs for businesses through personalized and timely employee engagement. Our client provides tools to analyze and differentiate the costs and quality of an extensive range of medical tests and treatments associated with employee health insurance.


To provide insights to the health benefits leaders, our client wanted to identify possible areas
of healthcare spend, generate personalized engagement campaigns automatically to connect employees to the most appropriate care or vendor program and report the impact of these campaigns periodically. To achieve this, Imaginea was commissioned to build a personalized, predictive analytics solution.


Imaginea proposed to build a predictive analytics solution on a SaaS platform with two different modules; one for employers who offer healthcare benefits, and another for their employees. This solution is data-driven, simple, timely and actionable. Salient features of this platform include:

  • Predictive models: The Predictive models are built, based on medical, pharmaceutical, internal, and social data. These advanced predictive models can automatically identify employees who are likely to need care. Based on predictions, employees are mapped to customized, cost-effective healthcare benefit or vendor program.
  • Multi-channel communication: Automatically deliver multi-channel, personalized healthcare campaigns to create engagement and awareness among employees.
  • Campaign dashboard: This dashboard tracks campaign engagement and impact on a daily basis. The dashboard data includes page views, vendor connections, provider selection, healthcare utilization, and much more.

Tech stack

How our solution helped

This increased user engagement in employer prioritized health campaigns, by personally targeting users, without affecting their privacy.

Overall approach

We built an Employee Health Opportunity Mapper (EHOM), which is the heart of the solution. At a very high level, the EHOM utilizes the user’s input data like medical, claims, search queries, demographic, social data and scores them against the Health Opportunity Model. The model predicts outcomes or identifies opportunities to engage with employees who are most likely to need care in the near future. Most of the data ingested by the EHOM is sourced from our client’s Enterprise Data Warehouse (EDW). The diagram below illustrates how our EHOM functions:

Here is our approach to specific functions:

  • Manage the data pipeline for EHOM: To manage the data pipeline for the EHOM, we used Luigi. It helps in managing complex pipelines of batch jobs and handles dependency resolution, workflow management, visualization, failures, and command line integration.
  • Run user score: Jenkins triggers a scoring run to identify opportunities. Luigi manages the dependencies between the tasks, for example sync in upstream data, get configuration data, create schemas, build aggregators, apply models, export results.
  • Historical analytics: Once the model identifies the health opportunities most suited to employees based on their attributes, the data is stored in the Enterprise Data Warehouse, built on Greenplum. We also store the user engagement data in the warehouse. This engagement data includes data for the number of users who have responded to the communication sent to them and the number of users who have performed a suggested activity. This data is then correlated with the Health Opportunity Model and presented as a chart on the client’s in-house reporting portal. This portal uses Postgres as its database and is built on Python and AngularJS.


The platform’s analytical engine runs simulations using millions of health plan pricing variables and identifies the best provider. This healthcare provider comparison became a core feature that helped our customer to differentiate themselves in the market. It offers:

  • A single, personalized view for employees to access all health plans, HSA, and plan status.
  • The intuitive, natural language search enabled employees to find the right information at the right time.

The capabilities to deliver insights powered by machines, resulted in improved quality of care and cost savings.

Talk to us