Using data to build better care systems
The following content is sponsored by Optum.
Ask most healthcare leaders what drives good patient outcomes and strong financial performance and you'll likely get the same answer: data, and not just any data. To get a full picture of their patients' health status, organizations need high-quality data from a multitude of sources, including claims, clinical, administrative and socio-demographic data banks. Such data can help payers, providers and the myriad support services that work with them understand how care is given, to what populations it is extended and how individual practitioners are performing.
It may sound impressive to say that your organization has access to terabytes of patient information, but without robust technology and smart people to manipulate it, that data is simply words and numbers without context. In a day and age when understanding patient risk is critically important, healthcare organizations must be able to cull robust data to build risk-bearing care systems and the financial models that will sustain positive patient outcomes.
And it all starts with quality data.
What is "good data?"
With respect to data quality, many factors come into play. Raw data from claims or from an EMR database are not suitable for analysis. Turning raw data into usable information requires preparation, including normalization and validation. Only then can an organization gain trustworthy insights from the information and put it to use in maximizing patient care, reducing risk and strengthening a business's bottom line.
While the concept of data quality is widely accepted, most healthcare organizations define "good data" in different ways. One common thread, however, is the overwhelming need to gather and analyze information from one end of the spectrum to the other — from all data sources and from all sites of care. To get a full picture of their patients' health status, organizations need high-quality data from a multitude of sources, including claims data, clinical data from electronic health records, administrative/abstracted data from facility information systems and socio-demographic data from public sources such as census data.
Human error is always a risk in data gathering and entry. It's not uncommon for patient data sitting within health system data marts to show men having babies, people born in 1776 and Daffy Duck coming to an emergency room.
And organizations must pay close attention to the sources of their data. "For example, while my company was cleansing data for a provider organization, we reviewed a lab feed that contained whole sections of lab values that could not possibly be human. As it happened, that lab was also serving veterinarians, and there was no designation for human versus non-human patients in the data," said A.G. Breitenstein, chief product officer for Optum. "That's an outlier example, certainly, but it's indicative of the fact that data can’t be trusted on its face — it must be analyzed and cleansed to ensure its quality."
More importantly, quality data must be actionable. "If data gathering is done simply for data’s sake, it is not worth doing,” said Adrian J. Rawlinson, MD, of Brown and Toland Physicians in San Francisco.
"Actionable data is useful clinical data that provides, for example, a pursuit list of high-risk patients or those likely to be admitted in the next six months," he said. "Anybody can create data or build dashboards and employ these tools. It is really a question of what you are going to do with it and how you are going to put it to best use once you have it."
One important use of actionable data is the development of accurate registries for care management. Registries, which are collections of health and demographic data for patients with specific health conditions, are traditionally built from claims data. Combining actionable clinical data with claims data provides organizations with a truer cohort of patients with the same disease. Without clinical data, registries will be incomplete. An analysis of Optum's clinical database of more than 40 million patients reveals that nearly 20 percent of patients with clinical evidence of diabetes lack a coded diagnosis of diabetes.1 That means one out of every five patients will not appear on reports from electronic health record systems by diagnosis code, on problem lists or in registries.
A true snapshot of risk
Generating and utilizing good data is merely a first step. The next step requires investment in advanced analytical systems that can provide accurate, timely and precise risk perspectives.
Baylor Quality Alliance, a clinically integrated organization of employed physicians, independent physicians, hospitals, and others associated with Dallas-based Baylor Health Care System, uses a blend of Optum analytics and home-grown systems for population segmentation, predictive modeling and performance measurements around quality and cost, according to Carl Couch, MD, president of Baylor Quality Alliance By focusing data analyses on specific functions, Baylor is able to quickly and effectively manage risk and make corrections to patient protocols where needed.
The Optum system utilized by Baylor applies four distinct components to maximizing all-source data analysis:
1. Integration of clinical and claims data across the continuum of care to give providers a complete view of population health
2. Better prediction of at-risk patients to reduce preventable costs via clinical analytics
3. Improved performance via deep comparative clinical benchmarks
4. Easy-to-use interfaces so non-technical people can interact without extensive training and support
"The fundamental assumption of analytics is you can't manage what you don't measure, and in healthcare we are particularly looking for performance measurement," Dr. Couch said. "When we look at Dr. A and Dr. B and Dr. C, we need to know why one of them has far better clinical performance or one of them has far worse financial performance than the others. That leads to discussions on what we need to modify."
Getting to the best protocols through robust analytics can mean combining different data sets within the analytic systems. For example, an organization can marry care management information and other analytics to see how it is improving over time.
"If an organization uses analytics to find a group of high-risk patients and assign some of them to intervention A and others to intervention B, in a year's time they can go back and bring those intervention variables back into the analytical platform," said Jeremy Orr, Optum Analytics' chief medical officer. "They can see which one worked better, and now have a tool that powers a continuous improvement process."
Powerful data = Better patient care
Physicians want to provide the best possible care to their patients. Although there is often wide variability in patient care patterns, the problem may be that physicians don't know how the care they provide varies in relation to peers within the same organization and in the industry as a whole. That's where quality data analytics can mean success within the realms of pay-for-performance and even fee-for-service.
When it comes to improving patient health, robust data needs to be applied to two related disciplines: identifying patient populations in need of intervention and identifying specific patient needs. When compared against evidence-based guidelines, these two disciplines represent true gaps-in-care management.
Analytic tools help organizations at a strategic level by identifying broad cohorts and segmenting those cohorts into targeted risk populations. For optimum usability, gaps in care should be stratified as age-, gender- and disease-specific. At a more tactical level, gaps-in-care management identifies specific gaps in the care of specific patients, which can be presented to physicians as a work list or within an electronic medical record environment.
Measure reporting, such as that done under Physician Quality Reporting System and Healthcare Effectiveness Data and Information Set, is more challenging. There are dozens of analytic tools that perform accountable care organization reporting, and others that specialize in PQRS reports. Efficiency comes with consolidating data analyses and reporting from one platform.
"Organizations want one solution to do ACO and PQRS reporting, plus other patient registry functions and pay-for-performance," said Dr. Orr. "They've been adding complexity to their IT environments for decades and now it's time to simplify." Dr. Orr believes CMIOs are hungry for a single data analysis tool they can use to parse all of the information from divergent sources quickly and efficiently.
Data and risk/value-based contracts
Having data and analytics to know where your organization stands amidst the larger universe is key, but it is sometimes just as important to have a good grasp on risk at the individual patient level. That's especially true when managing risk/value-based contracts within ACOs and similar organizations.
Risk-bearing providers are acquiring more advanced care-management strategies and an ability to build better predictive risk models for high-risk populations. Baylor Quality Alliance uses a predictive modeling system to better predict readmissions and segment populations to improve overall care management.
"In the United States, $1.35 trillion is the cost of care for five percent of the population, and they need special attention and resources applied to them," said Dr. Couch. "You don't know how to do that until you know who they are. That’s where good predictive models come in."
For example, Dr. Couch stated, an organization may be aggressively treating a cohort of diabetics. Without predictive modeling data, physicians may not know that half of them are depressed, which may lead to worse outcomes. Knowing this, he added, allows physicians and case managers to get those patients the mental health services they need. By doing so, those patients may be less likely to be readmitted.
"That is generally not dealt with in any other way other than the analytic systems that can help identify those patients and their specific medical conditions," Dr. Couch said.
Using analytics, care coordinators can reach out to patients to ensure they're seeing their physicians, taking medications, communicating with all of their care providers and understanding the different instructions they may get from multiple providers. Predictive modeling analytics allows organizations such as Baylor and San Francisco-based Brown and Toland medical group to reconcile discrepancies and enhance care.
While clinical analytics can help organizations identify clinical population and patient risk, organizations also need to know the details around their financial risk. This need calls for sophisticated predictive models that account for the health of an organization's patient population. Episode grouper technology — the same technology payers use to aggregate and assess data — is useful in this area. Using claims data to analyze past expenses combined with a disease knowledge base, groupers can apply advanced risk analytics to show organizations the cost of caring for a group of patients in the coming year. Such information is invaluable for budget and staff planning.
Build better care systems on better data
It all starts with data — real world data that can be applied to improve patient outcomes and remove unnecessary costs. For healthcare, real world data requires, at minimum, the integration of clinical and claims data to reflect what actually happens in health care environments. And by applying advanced analytics to such broad, inclusive data sets, providers will have a sharper lens with which they can analyze the inner workings of their systems. They can see what's working, what's not working, what it costs and whether it could be more efficient. Data allows them to better understand their populations, then tailor the care they provide to suit patient needs.
Applying advanced analytics to comprehensive data will reveal to providers things they don't know and help them more fully understand the risks and uncertainty associated with providing care. Analytics will help providers better calculate difficult decisions and prompt them to ask questions that they never thought to ask. Successful organizations will embrace the unknowns their data uncovers and ask what they can learn.
Providers want to know what they don't know, and they are using analytics to gain that knowledge — knowledge they need to improve, manage and succeed in today's dynamic, value-based healthcare market, as well as prepare for tomorrow's challenges.
1 Data from Humedica statistically de-identified common data repository