HIMSS 2014: Day 3 -- The State of Health Care Analytics
Day 3: Greetings from the land of HIMSS 2014 and Health Care Analytics!
Day 3 provided a wide array of uses for clinical analytics in a variety of settings: clinical integration, physician groups, ICD-10, surgical collaboratives, children’s hospitals, academic medical centers, and population health initiatives.
1. Clinical Integration to Support the Shift to Pay for Performance
The first presentation was an extremely interesting discussion from Phil Kamp, CEO Valence Health and Steve Cardamone, DO, MS, and CEO of Ohio Health Group. They shared their learnings in forming Health4, their clinically integrated network, and their strategy for garnering physician support for pay-for-quality initiatives and the impact of the network on improving quality metrics. Analytics ends up paying a very critical role.
They defined clinical integration as the agreement of independent physicians to practice medicine in an agreed upon manner such that they can collectively contract with payers. This requires participants to agree on common guidelines and metrics and have the ability to collect and provide performance verses clinical metrics to participating physicians.
Clinical integration allows more complete data, leading to better coordinated care across the continuum to deliver higher quality, lower cost medicine. Clinical integration is also a logical path along the evolution path to becoming an ACO.
Access to Data Is Key
Access to data was a key to Ohio Health Group success. They had a data warehouse which brought in data from physician offices, hospital data, lab data, payer/PBM data, and EMR data.
Among many lessons learned, these were the ones I found most important:
- Ohio Health Group designed its system with physician engagement as a leading priority
- “ACO” and “Bundled Payments” get the physician’s attention, but clinical integration is the mechanism to provide the flexibility to pursue those goals
- Relying on payer data alone is a short-term strategy only. You need more.
- Learning to manage as if you are under risk is not a bad thing because you probably will be done.
2. Leveraging Analytics to Capture the Full Benefit of ICD-10
Presented by Brian Levy, MD of Wolters Kluwer, this session focused on the difference of ICD-9 vs. ICD-10, and the analytics benefits to come.
Some people use the analogy of Y2K and ICD-10, but Brian felt the analogy doesn’t work for multiple reasons. ICD-10 will be harder.
- Y2K was a single date, but ICD-10 in reality is probably a 2-year rollout until stability achieved
- There will be an increased error rate through the entire transition time frame
- Although payers have a jump start, the predictions are still for a 100-200% increase in claims denials rates due to improper coding and other related errors
- And there will be significant over payment and underpayment due to errors.
- For those not careful, predictions are that 30 percent claims codes would be at risk for payment denial
However, we should not forget the longer term benefits
- Greater code specificity will increase our ability to report patient information and collect more granular information for future measurement and evidence-based research
- Increased documentation could lead to increased reimbursements rates due to the greater level of detail submitted
3. Improving Surgical Outcomes and Lowering Costs with Clinical Analytics
John Birkmeyer, M.D. presented some fascinating work done in collaboration with the University of Michigan and a broad collaborative of surgeons across the state of Michigan.
Surgery and specialty represent a huge portion of the total U.S. healthcare spend, and there is wide variation across health systems in Michigan in terms of utilization, per episode costs and efficiency, quality and safety.
Michigan has started a state wide set of collaborative quality initiative program, starting with a PCI pilot in 1992 to broad rollout in 2005-2006. They now have 15 specialty-specific programs, including cardiac surgery and PCI, medical oncology, bariatrics, breast cancer, prostate cancer, trauma/acute care, joint replacement, spine surgery and medical admissions. These programs share data and roll out best practices.
For example, when they agreed to collect data, they found the following $18K/episode variation in coronary artery bypass graft.
Example: Michigan Bariatric Surgery Collaborative
One bariatric surgery collaborative currently encompasses 33 hospitals, 75 surgeons, and 6,000 patients per year. One of the core accomplishments was putting together a clinical registry consisting of information from many more sources than just an EMR. These sources included:
- EHR (risk factors, treatments details, complications)
- Patient Portal (late complications, weight loss, comorbidity resolution, quality of life)
- Video Portal (peer skill ratings)
- Claims and Billing data (payments for facility, professionals, and ancillary care)
- Data Warehouse
The key conclusion is that the data from an EHR was important but certainly not sufficient. The collaborative needed data from many sources in order to put together a plan of action to improve care. This is the strength of a data warehouse – the ability to pull together many disparate sources of data (beyond the EMR) into a single source of truth.
Example: Reducing Venous Thromboembolism (blood clots) after Surgery
Dr. Birkmeyer shared some fascinating progress around VTE. At the outset of the project, one commonly accepted treatment for VTE included prophylactic IVC filters. However, data showed there was huge variation around hospital usage.
The Michigan collaborative wanted to drive a common standard but they knew that getting doctors to agree to treat care similarly would be the hardest part. So they started by collecting data. After some time collecting data and correlating the use of IVC filters and improved care, the group actually found a negative correlation. In fact, they found that in too many cases, the evidence showed that the IVC filters harmed patients. So they published that information backed by data, and usage dropped almost overnight. The insurance payment savings were so high in one year ($3M) that it literally paid for the entire program itself. Eventually sharing the common data and rolling out subsequent best practices helped reduce Michigan CTE rates by 50% (from .5% to .27%).
One other unique story was the group’s attempt to factor in surgeon skills. They were able to get every surgeon to send a video of themselves practicing. The video was edited down and made completely anonymous. Then they had a group of peers anonymously rate the surgeons’ surgical abilities. The findings were reported in the New England Journal of Medicine. The findings were “staggering,” showing that you can rate the skill of surgeons. They also found that the skill of the surgeon was a very significant predictor (highly correlated) with the patient outcome.
4. Texas Children’s Hospital
Next Dr. Charles Macias M.D., M.P.H. of Texas Children’s Hospital spoke. Health Catalyst has partnered with Texas Children’s Hospital on the organization’s data warehouse, analytics applications and clinical care process improvement projects for some time now. In fact, a significant detailed summary of the Texas Children’s Hospital clinical improvement success was recently published by CHIME.
In this presentation Texas Children’s Hospital shared its key strategies, results, and lessons learned.
Clinical Integration Strategy
Texas Children’s Hospital presented three pillars to their clinical integration strategy:
- Build a comprehensive, integrated, evidence-based quality improvement system
- Collect and meaningfully use data
- Implement an enterprise-wide data management infrastructure spanning multiple clinical and financial systems
Approach to Care Process Improvement
- Organizing permanent, integrated workgroup teams consisting of physicians, nurses, IT, quality and patient safety, quality improvement, clinicians, and business analysts that are responsible for a clinical program or clinical services over the long-term.
- Integrating critical elements of evidence-based practices into
- Establishing baseline measures, AIM statements with measurable goals and on-going review of results versus targets.
- Improved clinical care
– Decreases in LOS
– Decrease in readmission rates
– Decreased unnecessary procedures (i.e. chest x-rays)
– Millions in savings across several disease processes
- Reducing waste by systematizing reporting
– EDW reports cost 70% less to build
- Labor productivity tools allow global views for increased operational efficiency
- Best practices come from communities, not policy makers; these inevitably involve patients, doctors, nurses and others.
- The Pareto (80/20) principle is critical in the effort to prioritize. You have to be smart about your data collection
- Collect only what you need (they have a one-page rule)
- Incorporate data collection into existing clinical workflows
- The ROI on an EHR is difficult and not unlocked until you move to analytics enabled by a robust data warehouse foundation
- The correlation between cost and high quality is low
- Doctors will always find ways to tell you that your data is wrong; you need an analytics system that is agile and able to respond
- Data feedback must be timely or clinicians won’t act on it.
For those interested in reading more about Texas Children’s specific examples, please read the most recent CHIME report here.
5. Stanford Health Care — Improving Outcomes with an Innovative Approach to Population Health Analytics
The final presentation of HIMSS was also a Health Catalyst client — Stanford Health Care. Pravene Nath, MD, MSE and CIO, and Yohan Vetteth, MBA, VP of Healthcare Data & Analytics spoke about Stanford’s use of analytics in their population health initiatives.
Over the past 10 years, Stanford made multiple attempts to build an EDW. Originally they wanted to build it so they’d be able to answer any question, which led to a significant undertaking and effort to develop. This took a great deal of time with a substantial amount of money spent. At the same time, Stanford had difficulty engaging clinicians as the outcomes impact continued to be “out in the future.”
They also had a number of “ad hoc” efforts, which were quick but often with the tradeoff of not being scalable or not being as accurate as desired.
Eventually Stanford settled on an adaptive data warehouse approach, with the focus placed on clinical outcomes. They used only the relevant data to support their prioritized efforts rather than starting by trying to answer every possible question. This allowed the organization to start an incremental build adding new functionality along the way. It allowed the health system to quickly meet the needs of the most demanding constituencies in a timely manner.
- Stanford focused its efforts by prioritizing data-driven quality improvements efforts on the highest cost, highest variation projects.
- Their governance structure drives these quality improvement efforts through multi-functional, collaborative teams (clinicians, operators, informatics, quality, process improvement)
- They espouse an agile approach with frequent revisions to support user needs
- They need to engage with those requesting the data. “We have a 100% failure rate when someone just asks us for the data.”
CBA Governance Structure (Clinical and Business Analytics)
- Oversee the quality, access and standardization of enterprise assets (data governance)
- Provide oversight for resolving data stewardship conflicts
- Identify best practices in analytics to support a learning healthcare analytics environment.
- Integrated Multi-functional Collaborative Team structure
Stanford’s Population Health Efforts
Stanford showed its view of the paradigm shift to Accountable Care. So several years ago in April 2012, Stanford began the Stanford Coordinated Care Clinic. This initiative targeted patients with multiple chronic conditions and comorbidities, shown as tier 3 and 4 in the model below. It also included teams of medical professionals, healthcare coaches,and team care coordinators to help patents smoothly navigate their healthcare.
The Stanford Coordinated Care Dashboard
As a result of this initiative, Stanford shared its Coordinated Care Dashboard in several views.
Key Learnings and Conclusions
- Healthcare is ripe for analytics innovation
- EHRs are necessary but not sufficient; large gaps remain
- Designing an AICU and transforming primary care models is not easy
- There are shifting expectations of IT and analytics
- What does it mean to partner across different teams
- A multi-disciplinary approach is needed
- What types of people and skills are necessary to make this work
- How to we learn to ask better questions
- The clinical workflows that got us here will not get us there
- Build an adaptive data warehouse to provide value throughout the journey
- Focus on an agile, iterative approach to analytics
If you want to go fast, go alone. If you want to go far, go together.