Predictive analytics for value-based care success

Three years after the Medicare Access and CHIP Reauthorization Act (MACRA) was passed by Congress, the government is doubling down on its commitment to value-based care. As Health and Human Services Secretary Alex Azar recently told one industry group, “There is no turning back … in fact, the only option is to charge forward,” a sentiment frequently echoed by commercial payers.

For providers, this means steadily increasing accountability for the cost and quality of care delivery, as well as the introduction of meaningful positive and negative financial incentives based on performance.

But many practices are operationally unprepared to assume risk. They are still getting their arms around mastering the core requirements of value-based programs, including MIPS. In one recent survey, more than 70% said they had “a way to go,” while a majority foresaw the need for significant practice transformation. While this process has far-reaching clinical, operational and financial implications, a common denominator is the need to avoid flying blind.

Successfully navigating this environment requires medical practices to rethink how they leverage data for decision support – both what to prioritize and what to demand from their analytics. Fortunately, early pioneers in value-based care are generating best practices that peers can emulate.

Take, for example, the nearly 200 practices participating in the Oncology Care Model (OCM), a Center for Medicare & Medicaid Innovation (CMMI) advanced alternative payment model (APM) program launched in 2016 to provide higher quality, more highly coordinated cancer care at the same or lower cost. Practices receive a per member, per month payment for the duration of qualifying chemotherapy episodes to fund needed practice transformation, as well as a share of savings if they exceed cost reduction targets. Analytics have proven an essential component of their efforts along the way.

The path to predictive analytics

Practices that want to learn from the experiences of OCM leaders can refer to four common steps utilized in their data and analytics efforts:

1. Unifying EHR and claims data into a single patient view. In the fee-for-service era, practices tended to focus on EHR data as the primary source of clinical insight and claims data as its financial counterpart — but rarely did the two meet. Assuming risk for populations requires practices to have a longitudinal view of each patient’s health, across all comorbidities and care settings, as the greatest drivers of impactable cost occur outside of the office environment. Leading OCM practices have integrated patient data from across disparate systems, including EHR, practice management, payer claims, pharmacy, labs and more. They then unify and store it in a centralized “data lake” to enable calculation of total cost of care and serve as a single source of truth against which to run analytics.
2. Ensuring reliability through data curation. Integrating disparate sources is difficult enough, but practices must also ensure the resulting data is harmonized and provides a solid foundation for the practice’s value-based care efforts. Not only does this require technology interventions, such as identity matching and normalizing patient data, but often manual ones as well. Much EHR data, for example, is either unstructured — or structured but inconsistently entered. A vanguard of OCM practices have tackled this challenge with a dual approach. They have implemented health information exchange technology to handle data extraction and harmonization across disparate systems, but also employed skilled resources to complete and curate source data manually as needed. In addition, they have fostered education and cultural change among clinicians to maximize the quality of data from the start.
3. Generating actionable insights, not just reports. For maximum benefit, analytics must align directly with the measures and metrics embedded in payer programs, with a focus on surfacing those interventions to improve performance. For example, some OCM practices have adopted risk stratification tools that enable them to identify and target their high-risk patients, at both population and individual levels. This visibility enables them to take a series of preventative steps, from increasing office hours and navigation resources to employing targeted counseling to manage avoidable adverse effects.
4. Shifting from retrospective views to near real time and predictive analytics. As part of the OCM program, CMS provides participating practices with retrospective performance data every six months — too late for them to do anything about their results. Therefore, a vanguard of practices has implemented tools that allow them to monitor their performance against measures in near real time. Furthermore, they use sophisticated models to anticipate outliers in areas such as resource utilization and adverse events, which ensures they won’t occur in care management and other interventions.

Financial and administrative transformation: The next frontiers

While OCM practices initially focused on clinical-based analytics, they are beginning to consider bringing more actionable and predictive analytics to the financial and administrative realms as well. This is due to the changing nature of the physician practice revenue cycle.

For years, most practices embraced revenue cycle management (RCM) metrics tied to the amount and speed of fee-for-service collections. Analytics were enlisted to track first-pass resolution rates, payer denials and days in accounts receivable, for example. By contrast, value-based revenue streams can be more variable and difficult to predict, which can result in negative operational consequences if not closely managed.

The OCM, for example, includes upfront payments, shared savings arrangements and even potential clawbacks when CMS reconciles its definition of qualifying chemotherapy episodes with those for which a practice has already collected a per beneficiary per month payment. A new generation of RCM analytics, therefore, is emerging to span diverse contracts and drive accurate revenue projection across disparate value-based care activities on a payer-by-payer basis. The resulting insights should drive decisions such as whether a practice needs to take on stop-loss insurance to cover possible penalties or what percentage of revenue may not come in because of either internal errors or inaccurate reimbursement.

In the future, such analytics will also need to drive coordinated action between administrative and clinical practice leadership by predicting financial events with clinical dependencies and prescribing how to respond to them. For example, a forecast of underperformance on a payer quality measure may have a root cause tied to incorrect or insufficient documentation in a practice’s EHR — an issue that a lead administrator can raise to his/her physician counterpart.

It is worth remembering that carefully selecting which measures to report is another dimension of a practice’s value-based care success. The foundation of data and analytics described above can be used to illuminate those opportunities for highest performance in terms of cost-efficiency and quality improvements. It can also shed light on the quality of data available for accurate and sustainable reporting. Practices participating in MIPS can use these strategies to maximize the opportunity to achieve an exceptional performance bonus — a pool of $500 million awarded to practices that have most successfully progressed on their value-based care journey.

Conclusion

Value-based care has compelled practices to take on risk levels that were once the exclusive domain of payers. As a result, a new generation of sophisticated tools is needed — with analytics at the core — to rise to the challenge. At the foundation, practices should focus on:

1. Data integration across the range of healthcare systems, providing a holistic patient view
2. Data curation to ensure reliability
3. Tools that align directly with the quality and cost measures they are accountable for
4. Analytics that predict, not report, with near real-time visibility the performance and the ability to intervene on behalf of high-risk patients before the quality or efficiency of their care is affected.