How to Get the Most Out of Population Health Analytics


Post by Sumeet Shrivastava


VP Engineering for Population Analytics, Caradigm

It seems like there is a constant buzz these days around healthcare analytics and their role in population health. It makes sense because in order to effectively manage a population of patients, healthcare organizations need to gain a variety of different insights. For example, they need to understand the clinical and financial risk of their population, define cohorts and prioritize patients to be enrolled in care management, evaluate how they are performing against quality measures, view patterns in utilization for financial improvement, etc. Analytics provide these insights.   

However, in order to get the most value from analytics, it’s important to understand how analytics are integrated with other population health strategies. Analytics shouldn’t be viewed as an independent activity because population health is a connected series of activities impacting the entire organization and requires a broad set of capabilities including data aggregation, analytics, care coordination and patient engagement and outreach.  It is also important to consider end user needs such as making it easy to access, visualize and take action. Let’s explore a few of the key success factors needed for population health analytics in more detail.

  • Quantity and quality of data: These elements are essential to light up any pop health analytics scenario. Across many organizations, I have seen implementations get delayed due to “bad data”, i.e. data that is not ingested in the format that is required to produce quality analytics, and later requires additional work to cleanse the data to be usable. A data aggregation solution should be designed to handle bad data and provide tooling to make it easier and faster to consume data from different and often non-uniform sources. While the data from different systems will be inherently non-uniform, data ingestion tooling and software should considerably shorten the process. Quantity i.e. completeness of data is also important because incomplete data dilutes the value of the insights presented. Incomplete data can skew or inflate insights, which then can make it challenging to formulate actionable strategies based on the information. 
  • Easy and real-time access: Once data has been aggregated and cleansed, it needs to be easily accessible, timely and in a format that can be used by both applications and by non-technical analysts. If it takes days or weeks to access analytics, then that time lag can render the data out- of-date.  Creating consistent and reusable data models along with self-service tools better ensures that queries are simple, maintainable and timely. 
  • Actionable insight: For analytics to make the greatest impact, they need to be actionable and integrated into clinician workflows. For example, gaps in care can be surfaced in an EMR so that a physician can close them while still in the presence of a patient. Also, if an analyst sees quality measures below where they should be, the analyst should be able to drill down to the provider and patient level so that additional action can be taken (e.g. enroll patient in a campaign or the analyst creates a task to prompt a clinician to take action). Analytics should also be surfaced within care management workflows as part of a longitudinal patient record so that all members of a care team have a more complete picture of patients, which can then be incorporated into the plan of care.
  • Flexible visualization engine: A picture is worth a thousand words. Users need to able to visualize the data in various formats – whether it is graphs, plots, trend lines, etc. An analytics engine should support different forms of visualizations that give users flexibility to derive value. 
  • Advanced analytics: To evolve beyond retrospective reporting and receive greater value from analytics, organizations need to leverage predictive and prescriptive analytics. Predictive analytics are essential for prioritizing resources as they forecast clinical risk, identify cost savings opportunities, likelihood of readmission, and can even identify which patients are likely to comply with a plan of care. Based on these predictions, prescriptive analytics can go one step further by suggesting interventions and identifying what actions should be taken to improve care for the patient.

Caradigm has taken a holistic approach to population health analytics since our inception.  We designed our enterprise data warehouse, analytics applications and workflow applications to work together, which makes each individual application more effective. Our analytics offering is unique because it can leverage virtually all of an organization’s data from disparate systems in real-time, apply best-of-breed algorithms to that data to derive insights, then surface that information directly within clinician workflows to drive action. If you’d like to discuss pop health analytics more, then please send us a note here. I look forward to continuing the discussion on analytics in an ongoing series of posts that will be coming from the Caradigm team.