What Are Modern Data Platforms, and How Can They Boost Healthcare Agility?

Understanding the Modern Data Platform for Healthcare

Atlan defines the modern data platform as “the technological base on which data tools and applications connect or sit” and describes it as “the operating system of the data stack.” Databand, a subsidiary of IBM, notes that the modern data platform “should provide a complete solution for the processing, analyzing and presentation of data.”

Olenik cites two primary goals for the modern data platform. One is eliminating barriers to access, especially for external data sources such as insurance claims and patient-reported outcomes. The other is ensuring that data, regardless of its source, is ready to be used.

Unless data is clearly defined using taxonomies or dictionaries, he says, “you’re going to be dealing with problems right up front. You need to set standards related to business processes, quality, access and security. If different standards are in place, that only leads to difficult decision-making.”

While some health systems are comfortable doing this work internally, “it’s not a core competency of all organizations,” notes Jonathan Shannon, associate vice president of healthcare strategy for LexisNexis Risk Solutions. “Those that think of themselves as care settings first are going to lean heavily on their vendor partners.”

EXPLORE: How the cloud improves data sets for real-world evidence in clinical trials.

Achieving Healthcare Agility with Modern Data Platforms

Agility is one of the core benefits of implementing a modern data platform and combining it with data governance best practices. Fast access to accurate data leads to more informed decision-making across the organization, as data scientists are no longer the only ones looking at the data. Business leaders can pivot as necessary, while clinical teams can gain a better understanding of a patient’s condition and adjust treatment plans accordingly.

The modern data platform also contributes to efficiency and cost savings, adds Priya Krishnan, head of product management, data and AI at IBM. “There’s no more duplication of data, and there’s higher-quality master data,” she says. “Without the manual effort of preparing data for regulatory compliance, your data stewards can be doing more meaningful work.”

The continued evolution of automation technology will open additional opportunities for data discovery, indexing and classification, Olenik says, to the point that these tasks may no longer require a background in data science. “Ultimately, self-service analytics is the goal, as data scientists aren’t plentiful.”

Click the banner below to learn how a modern data platform supports decision making.

The Role of DevOps and DataOps in the Modern Data Platform

The principles of DevOps and DataOps — which aim to integrate IT operations with software development and data management, respectively — align well with the modern data platform.

“It will be transformative,” Krishnan says. “You can create high-quality data products that are consumable by different business units and applications, and it’s a repeatable process.”

Olenik views the implementation of these best practices as defining “the cursory work of what we’re trying to do” (DevOps) and using the available data to put it into practice (DataOps). When data is high-quality, accessible and secure from the beginning, it results in better products.

“It’s beginning with the end in mind,” he says. “You need to think about the data requirements, and you need to think about data integrity and security and business rules. Otherwise, you won’t achieve your end goals.”

This line of thinking shouldn’t be limited to the IT department, Olenik says. The operational perspective is important — and so, too, is the input of the “heavy data users” who can highlight data-related issues or define the business rules that are written. This also helps organizations have more informed conversations with their vendor partners, particularly when it comes to quality and integrity controls, he says.

Elements of the Modern Data Platform

There are multiple elements to the modern data platform. Each plays an important and unique role, and each comes with its own advantages and drawbacks.

Data Ingestion

This refers to the process of moving data from its source to a centralized storage system for future use. Data ingestion is most effective when it occurs automatically, either in real time or in batches, as this helps to ensure that no data is lost in transit. It’s important to get this process right, as applications downstream for reporting, record-keeping, analysis and more rely on the accessibility and accuracy of data ingested from its original source.

Data Pipelines

In the past, ingesting data also meant reformatting it before storing it. Cost was the primary motivation for this process, known as extract, transform and load (ETL); given the resources required to both store data and run analyses, it was critical to get data in the right format before delivering it. As computing costs have decreased, data pipeline products have emerged to load raw data for interim storage, transform it and then store it at its destination. This is less disjointed than the ETL process while also being more efficient, thanks to the elasticity of cloud computing.

Data Warehouse

Organizations typically store data that’s structured, formatted and ready to be used in a data warehouse. Like a commercial warehouse, a data warehouse is intended to be organized so that anyone with access rights can quickly find what they need for reporting and analysis. In other words, the data in a warehouse serves a specific and clearly defined purpose, and in some ways a limited one, given that the amount of unstructured data in healthcare is growing.

Data Lake

A data lake, meanwhile, serves as a repository for raw structured and unstructured data after it’s been ingested and reformatted but before it’s been put to use. As with a data pipeline, a data lake takes advantage of the low cost of cloud computing (in this case, object storage). Though the data lake has a less formal structure than the data warehouse, some form of organization is necessary. Likewise, IT teams still need to set security protocols and determine access rights for a data lake.

Click the banner below for more insights on the growing importance of data in healthcare.

Data Mesh

Organizations or business units that feel constrained by centralized or otherwise siloed data systems may turn to a data mesh. In this data architecture, the so-called “domain data owners” are responsible for ingesting and formatting their data, though the organization as a whole still handles data governance and other policies and standards. A data mesh lends itself to a flexible self-service model, giving individual teams greater control over their data while reducing the computing load forced upon the centralized data platform.

Data Discovery

Data discovery refers to the process of collecting and evaluating data from disparate sources, primarily for the purpose of identifying trends. Given its close association with business intelligence, data discovery involves visualization in addition to analysis. The modern data platform is a key enabler of data discovery, as it eliminates the need to manually aggregate data from silos and normalize it prior to analysis.

Putting the Modern Data Platform Together

For Krishnan, there are four underlying principles of modern data platform technology: scalability, integration, data quality and governance. With these in place, organizations are better positioned to innovate, and to use that innovation as a competitive advantage. “It can be a game changer,” she says.

With so many modern data platform elements to consider, and with healthcare’s operating margins and IT budgets limited, organizations must think carefully about where it makes sense to add another layer of technology, says Shannon.

There are trade-offs to consider, and they will likely depend on how advanced an organization is in its data management strategy. “Is your primary need to ingest and manage data,” he says, “or is it to pull data out to marketing, operations and clinical stakeholders?”