Improving care with analytics

May 13, 2016
By Neil Smiley

Most hospitals have been leveraging analytics to address a variety of operational needs from procurement to revenue cycle management, and patient risk analysis. However, value-based payment models are blurring the traditional domain boundaries for analytics. Hospitals now need to link data sets that have not previously been brought together as hospital profitability and clinical outcomes are increasingly interdependent. Hospital data alone is also insufficient. Analytics need to extend visibility beyond the walls of the hospital to include data from payers, physician groups, post-acute care providers and community-based partners.

Beginning in April 2016, CMS moved to bundled payments for hip and knee replacement surgeries for approximately 800 hospitals that reside within one of 67 geographic areas. The bundled payment includes the hospital procedure and all costs incurred in the 90 days following discharge. Affected hospitals are now thrust into the role of a network convener, responsible for cost and outcomes for the entire 90-day episode of care. It is widely expected that CMS will continue to expand the bundled payment program to include other geographies and procedures. Other payers are following suit.

Bundled payments and other value-based reimbursement models require hospitals to embrace a comprehensive and integrated approach to analytics in three key areas:

• Network Design
• Network Management
• Network Interventions

Network design
In a fee-for-service model, hospital financial responsibility ends at discharge. However, with new value-based reimbursement models, hospitals are on the hook for care episodes that typically extend 30 to 90 days after discharge. To make matters more challenging, most hospitals do not have their own post-acute care resources. For successful outcomes hospitals will need to collaborate with financially independent post-acute care providers. And yet, hospitals will increasingly bear responsibility for both financial and clinical outcomes.

Hospitals need to start by getting a handle on the referral patterns and performance of existing post-acute care partners, particularly for conditions and care pathways involved in value-based care reimbursement models. Most hospitals have found that relying on self-reported data from post-acute care providers is usually insufficient, as there are too many data visibility gaps and too much inconsistency in measurement methods.

Public quality data, such as the CMS websites for Nursing Home Compare and Home Health Compare, are too blunt of an instrument to guide network selection. Hospitals need to develop their own ability to evaluate cost and quality performance of network partners, and where necessary, narrow their networks to those providers that are able to meet performance criteria. CMS and other payers are beginning to make claims data available for patients in risk-sharing reimbursement models.

Whether under bundled payments, ACO or managed care models, hospitals must be ready to leverage claims data to develop standardized, risk-adjusted measures to inform post-acute care partner selection and performance. On an on-going basis, claims analytics can help pinpoint sources of variation in cost and quality to enable the development of effective intervention strategies and continuous improvement programs.

Network management
While claims analytics can provide a big picture overview of performance, claims data typically lags by several months. It’s difficult to steer process improvement using only a rear view mirror. To effectively manage patients across a network of providers, real-time analytics are also needed. Hospitals will need data sharing agreements with post-acute care network partners, and an ability to integrate data from multiple systems.

With data sourced from different organizations, a Master Person Index (MPI) will be needed to link patient and provider encounters across systems, and care settings to construct a longitudinal view. Once real-time data has been matched up, more comprehensive analytics can be developed that leverage data from network partners. Real-time patient tracking can be linked with costs from historical claims data to establish predictive cost models that help network providers anticipate episode costs and preemptively intervene with high-risk patients.

Network interventions
As patients move through the care continuum, data markers need to be continuously monitored in real time to identify high-risk patients that need more intensive care. Predictive analytics can identify patients of rising risk to match with appropriate interventions. With an effective feedback loop, advanced machine learning algorithms can optimize risk models over time to improve accuracy as hospitals expand their access to new data sources and gain experience.

About the author: Neil Smiley founded Loopback Analytics in 2009 to deliver an advanced Software-as-a-Service platform health care providers can use to prevent costly readmissions. The Loopback Analytics team currently works with the largest pharmacy, hospitalist group, health system, payer, and senior housing provider in the nation, providing proven intervention solutions that improve clinical outcomes and reduce the total cost of care.