Provider attribution, the process of associating outcomes to the care of individual providers, is a critical component of an effective performance improvement initiative. Hospitals can realize big gains using data to drive performance—but doing so requires understanding how providers, patients, and diseases map to measurable outcomes.
These relationships may seem hopelessly complex, but big data and advanced analytics have promised to explain the connections. Unfortunately, it has proven difficult to translate these theoretical advances into physician practice. A 2015 study by Jvion found that only 15% of providers are using any kind of predictive analytics. According to the Harvard Business Review, the difficulties have not been with the analytics, but with getting buy-in from decision makers.
As the effects of value-based payment models are felt, more and more hospitals have an urgent need to overcome these challenges. As of 2016, 50% of overall reimbursement is tied to value measures, and hospitals can receive up to a 34% adjustment to annual medicare reimbursement based on provider quality and performance metrics. Government initiatives have increasingly focused on holding hospitals accountable for the success or failure of the entire system, and organizations are feeling immense pressure to efficiently improve outcomes.
With so many factors influencing health, and so many components of the healthcare system to analyze, it can be hard to focus initiatives where they will be most effective. Physicians are just a portion of the system, but we rely on them for a disproportionate percentage of the important decisions. When hospitals have successfully implemented data-based performance improvement programs for physicians, they have proven effective.
But what makes some initiatives successful while others are ignored?
Fair and accurate metrics are the key to physician engagement and to making data actionable. A number of organizations have found that when physicians feel they are being judged based on unfair metrics, they disengage with improvement initiatives. Insights that focus on factors that a physician can control are much more actionable, according to the Society of Hospital Medicine.
How to build a metric that drives improvement
An effective metric will...
- measure outcomes that are meaningful to providers and aligned with administrative goals
- be understandable by its consumers
- be appropriately risk-adjusted so that comparisons can be made across clinical contexts
- reflect the degree to which a physician could influence each case's outcomes
Measuring and reporting alone are not enough. A provider’s ability to affect outcomes changes on a patient-by-patient and case-by-case basis. The critical step—adjusting metrics based on what is fair for comparison and actionable for the physicians—is often overlooked.
How are metrics attributed today?
Many attribution models have been developed, but most organizations utilize methods based on reporting convention rather than accuracy. Attribution was historically a billing requirement, and focused more on legal accountability than on quantifying physician influence. Two basic types of methods are common: single attribution and multiple attribution.
- Single attribution is the assignment of a patient to one provider for accountability. This model was created many years ago for billing and legal purposes. Despite its inability to account for outcomes influenced by multiple providers, it is the most common method in use today.
- CMS uses a multiple attribution model, where visit-level stats like length of stay (LOS) are associated with both the discharging and attending providers. Other metrics are attributed based on the percentage of care billed.
In either case, accountability is all-or-nothing. These methods poorly reflect the collaborative nature of today’s medicine—where a spectrum of influence is distributed asymmetrically across the care team.
Metrics that reflect performance
Recognizing the value of proper attribution, healthcare organizations have developed techniques to quantify provider influence over outcomes. For instance, the Responsible Provider Algorithm developed by the University of Minnesota assigns a single provider to inpatient accounts using clinical notes, procedures, and provider specialty. The algorithm has helped them to improve physician quality measures and performance reviews.
Johns Hopkins developed a method for attributing patient‐level metrics to rotating inpatient providers, assigning multiple providers a portion of the metric, weighted by the percentage of charges entered by each provider. They have seen dramatic improvements in patient satisfaction scores and quality metrics since implementation.
These success stories demonstrate the potential of accurate attribution, but there is much more room for improvement. In many cases, the algorithms require manual setup and maintenance, or they focus on only one or two factors. Not only do these methods lack scalability and transferability, they fail to capture the important signal that points physicians toward cases with outcomes they can influence.
Research has shown that metrics must account for the overall hospital environment when assessing provider performance, leading CMS to propose changes to their attribution model. Further research has improved the accuracy of provider influence calculations and shown that meaningful insights can be drawn even from small data sets.
By combining the principles of partially-weighted attribution with advanced analytics techniques, it is possible to create an automated solution to quantify provider influence on individual outcomes. Organizations that solve the attribution problem will be able to drive performance improvement—allowing them to stay ahead of reimbursement trends, reduce costs, and improve outcomes.
Creating metrics that are fair, meaningful, and actionable is the core of Agathos’ analytics model—and its company vision. Our customers have already seen increased physician engagement and improved outcomes, and there is much more progress to be made.
If you are a researcher or a hospital interested in hearing more, we would love to discuss.
About Author
