Effective Data Mastering at Scale for Managed Healthcare Organizations

The managed care business is complex. Tamr can help with that complexity.

While we may try to explain managed care as the interactions among patients, providers, and payers, industry experts know there are much more layers and vehicles involved. The below tables show a sample of the business, stakeholders entities, and levers involved in the ecosystem.

When analyzing managed care data and building models to inform business strategy, life science analysts need to, among other things:

• Identify relevant business questions and processes
• Understand the key entities involved in the analysis
• Find and access datasets containing relevant data metrics and information around key entities
• Resolve data issues such as discrepancies, inconsistencies, missing data, duplicates, misclassification, and other issues that may require the help of subject matter experts
• Build an analysis based on the collected information
• Do significant re-work if assumptions or requirements change at any point while building the analysis
• Repeat…

Despite large volumes of available data from vendors, one of the most time-consuming challenges is still the data preparation before the analysis can begin. Depending on the scope of the analysis, a data prep tool or even database queries may be sufficient in supporting data analytics. However, those approaches do not scale.

Instead, customers like Sunovion have used Tamr to build a maintainable reporting layer that aggregate, master, and standardize both internal and external data around hierarchical entities such as corporations, payers, sub-payers, health plans, etc. Tamr’s agile machine learning approach creates an incremental data hub over months, rather than years, and provides companies with significant leverage over their competitors in generating on-demand analysis for sales and marketing strategy, as well as managed market / market access analysis.

Critical analytics for sales, marketing, and managed markets / market access

Company executives and decision makers need vital insights around the managed care ecosystem for their business; despite growing volumes of data sources readily available, timely and accurate analytics continue to be a costly challenge towards building a robust data-to-insights cadence.

Navigating relationships in the managed care market has traditionally been highly manual, slow, and not without a degree of inaccuracy. Below are examples of familiar business questions and analytics that should be easily produced in days, but in reality may take weeks to months, depending on the maturity of a company’s data management infrastructure.

At a specific geography, who are the top prescribers writing scripts for a specific drug and payer? What are the associated key performance indicators?

Which metropolitan statistical areas are best to target deployment of coupons for medicare part D patients?

How do the top metropolitan statistical areas compare with each other and the nation related to rejection rates and market share?

These types of analytics support business strategy and decisions that may have millions of dollars on the line. How quickly and with how much confidence such analytics can be produced is critical for data-driven organizations making huge decisions while navigating and negotiating relationships within the managed care ecosystem.

Current approach to commercial analytics lacks data mastering

Efficiency and accuracy in data analysis for enterprise decision making has long been a point of focus for data-driven organizations. Traditionally, life science companies have tackled data initiatives in various ways, including:

• Internal analysts – hiring analysts to execute data analysis of healthcare data to solve specific business questions over 1-2 weeks. This is a highly manual process bound by internal resources and expertise.
• External third-party contractors – hiring external contractors on a project-basis to answer large business questions over a 1-2 months. Typically locks the company with the same contractor over several projects due to inefficient knowledge sharing and repeatability.
• Traditional data warehouse – hiring external contractors to build a data warehouse for business intelligence reporting. This approach typically takes months to years (if completed) before reliable reporting is in production and requires both SMEs and developers to constantly dive into rule logic to QA and manage the integration of data sources.

Any of the three traditional approaches mentioned above work – but each has their drawbacks and limitations. Most importantly, these approaches do not scale in the long-run, while incurring increasing large costs over time.

This problem is not limited to the life science industry, and we have seen customers use Tamr to break themselves out of a dysfunctional or slow data-to-insight cycle in favor of something that is more agile and scalable.

Tamr’s agile data mastering approach is powered by machine learning

Tamr uses a machine learning approach to combine, master, and classify data. There is a lot of content describing Tamr’s technical methodology for agile data mastering, and I will not go into too much depth here.

However, I have provided an example below to illustrate how Tamr can be used to combine two disparate commercial datasets with different data standards into unified datasets for specific entities. Tamr focuses on curating data around specific entities based on data coming from the disparate sources that are ingested into the platform, and generates golden records of discrete, reconciled entities that analysts can immediately use for analytics.

As shown above, disparate datasets processed through Tamr help generate unified data models around specific entities, thus removing the need to spend hours reconciling data inconsistencies with subject matter experts.

Instead, a streamlined flow can be created to continuously load multiple sources incrementally into Tamr and provide a sustainable pipeline for your data reporting infrastructure such as the below data flow:

Reducing time spent on enterprise data wrangling, and spending more time on generating impactful analytics through innovative AI, data science, and modeling techniques will be critical for life science companies to gain a competitive edge in their commercial business in the long run. However, life science companies must beware of placing the proverbial AI “Cart” before the Data “Horse”.

Conclusion

The ecosystem surrounding managed care and healthcare is complex. As the volume of data being collected for this ecosystem is only going to continue to increase over time, the issue around data variety will always exist. Building out a robust DataOps infrastructure and a focus on data mastering to manage this data with best-of-breed technology such as Tamr is the only way for life sciences to maintain their edge as we continue to be dependent on data to drive business success.

‍