by Paul Bradley

Bringing the agile development model to the healthcare revenue cycle

Aug 23, 2016
AnalyticsSoftware Development

A look at implementing predictive analytics


At the Snowbird ski resort in Utah in 2001, a group of 17 software developers gathered together and hashed out a manifesto.

Fifteen years later, organizations continue to benefit from the principles outlined therein — the Manifesto for Agile Software Development lives on.

Yet rather than dwell on the founding narrative — spritely told, and worth a read — we should perhaps look instead to how the principles of agile software development can be shared between software companies and the clients those companies serve — how software companies can help their clients find opportunities to implement the principles of agile in their own organizations.

A fitting example exists in healthcare, where there’s not only the business office side of the clients to consider, but also the clinical leaders and faculty — and in-house IT professionals, too.

Let’s take a look.

Implementing predictive analytics in the hospital and health-system setting

The principle of “Close, daily cooperation between business people and developers” (one of 12 principles of agile development) must exist in the client-vendor relationship. Greater results are achieved through collaboration, because the agile model keeps momentum moving forward continuously through structured, incremental deliverables.

An implementation of predictive analytics at a large U.S. healthcare system encompassed these three phases:

  • Data validation: Rigorous testing of data attributes in the source and ancillary systems.
  • Reporting analytics: Creation of comprehensive analytics and reporting data warehouse.
  • Customized workflow: Operationalizing actionable analytics into work queues.

The rigorous data validation is vital for a few reasons. First, because healthcare data is disparate and diverse — it exists in multiple forms, often in multiple systems. Second, to state the obvious, meaningful data aggregation isn’t as simple as dumping all the data into one place — the technical bringing-together of the data must be done correctly. Where there are inconsistencies, there must be adjustments — or the end result will simply be a proving of the maxim, “garbage in, garbage out.”

With the data layer validated and established, we built the analytics layer. Based on collected patient-level data elements, we focused on predicting the probability that:

  • A charge is missing — with the goal of identifying charge-capture leakage yet not creating manual exceptions.
  • A patient will pay their portion of the bill — so that the client can engage in more efficient and targeted followup.
  • A patient will readmit — so that an alert can be automatically raised (before the patient is discharged) when there’s a high probability of unplanned readmission.
  • A DRG was assigned in error — so that the right claims can be automatically routed to coding/CDI teams for followup before the claim is submitted.

Our team’s agile development model centered around the following core client business goals:

  • Uncovering account populations that require intervention — while simultaneously identifying root-cause process improvements.
  • Quickly responding to issues and high-risk populations.
  • Enabling end users at all levels to not be overwhelmed by the volume of data and enhanced access to it.
  • Structuring small, consistent steps to rapidly and continuously drive change.

This approach, in turn, empowered the health system’s management and staff to proactively identify issues in their areas of responsibility and bring solutions to the leadership team. Just as important, the agile development methodology bled into their efforts to improve revenue-cycle performance. The principles of agile imbued into the software development were mirrored in the efforts undertaken using that software. For example, the principle “Simplicity — the art of maximizing the amount of work not done — is essential” was reflected in the newfound ability of health-system staff to work by exception. They no longer had to do the work of identifying which accounts needed attention — nor did they have to attempt to manually determine which of these accounts were the highest value (and most likely to yield a positive outcome). The “work” was done by the culmination of a clean, verified data warehouse, predictive analytics to identify at-risk accounts, and routing them to the staff best suited to intervene and resolve — which minimized the amount of work required by end users of the software.

The results

The principle of “the most efficient and effective method of  conveying information to and within a development team is face-to-face conversation” led to improved communication and coordination across all revenue-cycle related departments. Automation ensured reduction in manual, non-value-add tasks. At-risk and outlier populations were effectively prioritized for proactive and immediate followup. And the organization sustained permanent improvement in the quality and integrity of their data — both in their predictive analytics solution and in hosted, installed systems.

They were also able to simplify their technology portfolio, removing bolt-on solutions that no longer added value or provided accurate information. One way to think about this is that they maximized the amount of vendors they don’t work with — by focusing on simplicity that delivers the full functionality they need, rather than multiple vendors who all provide some sliver or approximation of what they hope to have.

This may all feel a bit academic — a data scientist suggesting that hospital and health system leaders need to take a page from the software development world. But what matters isn’t where the ideas came from, or how they were originally applied. For business leaders, clinical leaders and technology leaders, what’s important is that these principles can help ensure stronger financial performance, better use of staff time and more targeted attention paid to the patients and patient populations in greatest need of proactive intervention — all ingredients that help strengthen a healthcare organization’s bottom line, providing the fuel to improve patient health and outcomes for years to come.