Home  |   Subscribe  |   Resources  |   Reprints  |   Writers' Guidelines

April 2017

Digging Deep for Revenue
By Susan Chapman
For The Record
Vol. 29 No. 4 P. 10

A myriad of factors stand in the way of health care organizations using data analytics to boost reimbursement.

As hospitals transition to a value-based reimbursement model, data analytics may play a leading role. However, with an overwhelming wealth of information being generated each day, key decisions must be made to ensure the captured data are truly of benefit.

Capturing the Proper Data
Jeff Currier, data analytics manager at Availity, says determining the type of data that must be captured depends on the analysis. "There are many types and sources of data to be considered. For instance, claims data, which details the types of conditions and diagnoses, will give a picture of the patient's health history," he says. "Data can also be structured, like that in the EMR, and unstructured, like patient engagement metrics about hospital admission and emergency department experiences, and how those experiences play into a positive outcome for the patient."

In general, two types of data from EMR and claims data sources measure quality, says Jason Burke, system vice president and chief analytics officer for UNC Health Care. "Clinical data include answers to questions such as, 'Are diseases being well managed? Are the right treatments being prescribed? Do patients have to come back to the hospital 30 days after discharge?' Process data, however, pertain to whether or not patients have been screened for certain diseases and health risks, like diabetes and smoking," he says.

For data to be analyzed properly, the information, regardless of its form, must be accurate from its point of origin. Because caregivers need to have accurate patient information at their fingertips to dispense effective care, structured data are crucial. "When providing immediate care, hospitals don't have the luxury of having data crunched. The data need to be structured at the point the information is gathered in a clinical format; any analytics has to happen after that," says Jay Anders, MD, chief medical officer at Medicomp Systems. "They need to collect pertinent patient data—the patient's condition as recorded by a clinician or a nurse. And then those care providers have to be able to act upon it in real time."

Data must also be usable over the long term. "From my perspective, data need to be longitudinal in addition to episodic," says Dorrie Guest, managing director at Deloitte Consulting LLP. "In a population health approach, you're not just focused on one patient, one episode. You're trying to keep people from becoming patients in the first place. The data, therefore, have to provide actionable insights on costs, service, utilization, quality, outcome, and low- or high-risk patients over time. Practicing medicine is a team sport for those with chronic conditions, which involve multiple care providers. In that scenario, the data analytics lens must turn its focus on the individual's pathway to care that is high quality yet low cost over the long haul."

Chronic conditions, such as congestive heart failure, require that a number of health care professionals engage in the patient's care plan. As a result, Guest says effective analytics must follow the care process from the patient's point of view to find the optimal care pattern at the lowest cost.

Generation Speed
Because data analytics requires a range of information—from lab information and pharmacy data to claims and cost information—how quickly it can be generated varies. "Most organizations have requirements on completion of documentation in the EMR. Those data can usually be obtained within 24 hours," says Mary Branagan, MS, RN, CPHIMS, a senior health care consultant at Change Healthcare. "Some data are real time and can be shown in the user interface immediately after they're entered. Other data, such as claims data, are much more variable."

Tina Foster, vice president of business advisory services at RelayHealth, says the staggering amount of available data can be overwhelming. "Health care data is at more than 150 exabytes, an incomprehensible scale, and it grows at 48% annually, so the generation of data is not the problem—it's the generation of information that becomes the crux of the matter," she says. "The data are everywhere, but how quickly can they turn that into actionable insights? It depends on the computing horsepower to acquire it, mash it up, and make sense of it. With privacy rules—who can have access to what information when—this process gets even more complicated. How do we harness the data out there? How do we aggregate it and enhance it with sophisticated logic across all these sources to get an extensive view of one person or population? That is the challenge. It becomes about investment in infrastructure, computer horsepower, and competency in the people looking at the information."

Foster says the speed and accuracy at which information is generated and captured depends on the three Cs: curation, culture, and competencies. "Culture is how your organization embraces the process," she says. "Staff have to look at information as a strategic asset. We have to have competency, the right types of people with the right data and analytical skills. Curation drives the confidence, having faith that the right data are coming in consistently and without gaps to actually translate to highly reliable information that will inspire clinical confidence to drive transformation."

Difficulties in Accessing Claims Data
Highly standardized claims data are key to the analytics puzzle and how it applies to reimbursement. However, stumbling blocks such as the age of the data, their retrospective nature, and a lack of data-sharing agreements can present serious challenges. "It's much easier to get in-network data and more difficult to access out-of-network information," says Billie Whitehurst, vice president and general manager at McKesson Connected Care and Analytics. "Data-sharing agreements are important. The holy grail to providers would be to enable access to data across the continuum, including from extended care organizations such as home care agencies and skilled nursing facilities. This granularity would help providers truly provide value-based care."

Unless they are in a risk-bearing/value-based care arrangement, providers typically do not have access to a full set of claims data for a particular population. "As more and more risk-based arrangements are made, you need to have the full complement available," Guest says. "Medicare and other payers are increasingly making that information available. The problem is typically latency with claims data being available at best six months to a year after services are rendered. The really innovative analytics pair clinical data, real time, with claims data."

Branagan says claims data have several dependencies, including the need for coding, that can create delays. "Proper coding is key to reimbursement and significantly impacts value-based payments. Coding has a critical impact on the ability to define risk, which impacts the level of reimbursement," she says. "With the implementation of ICD-10, coding has become more granular, and providers are still in the learning curve. Many providers do not have expert coders in their offices. Some EMRs have coding components, but expertise is required to program the systems in such a way that optimizes the process."

The manner in which claims are submitted can cause delays, Branagan notes. "There are a variety of ways a claim can get to the payer and that sometimes involves using a scanning company and/or claims clearinghouse. The claim must meet CMS [Centers for Medicare & Medicaid Services] claim processing and billing guidelines in order for it be accepted and adjudicated by the payer, and that is not always a one-step process. Once the claim is processed, the data can be used for analytics," she says.

Different Data Formats
"Using data to make better decisions is exciting, which has been tempered over time," notes David Delaney, MD, chief medical officer at SAP Healthcare Sector. "I think differing data formats, especially structured and unstructured, are one of health care's most thorny challenges."

Interoperability among systems of different ages and operating platforms can also be problematic, a challenge for which Guest sees machine learning, or artificial intelligence, as a viable solution. Laboratory, pharmacy, and X-ray information are examples of areas in which disparate systems have been able to aggregate information for analysis. However, unstructured clinical encounter data will require advanced technology such as machine learning and natural language processing to make it useful for aggregation, analysis, and reporting purposes.

Still, while the health care industry continues to work on interoperability, there are effective processes that can be drawn on to help facilitate that transformation. "Any standardization efforts must start with data governance," Branagan says. "An organization must define the data elements across internal departments. This is the first step in standardization. It is surprising how data definitions vary across an organization. Financial definitions differ from clinical definitions, and even some of the most basic data elements can be retrieved from databases in ways that give results that do not match. This can cause widespread confusion and result in poor data quality. It's imperative to have a solid process for data standardization in order to protect the integrity and validity of the data to do effective data analysis."

According to Currier, connecting with those areas in-house that have already developed data in a standardized format is the best way to overcome the challenges of data being in various formats. "Formats aren't necessarily the biggest problem, just a challenge. With the right partnerships, they don't stand in the way—they just slow it down," he says. "We don't have a one-size-fits-all standard in the industry at this time, and we may not. There are some challenges in terms of accuracy and dependability, so there needs to be a well-documented process as to how the information got from one partner's system into another."

Brian O'Connor, vice president of enterprise technology at Evariant, believes a big data platform is the most appropriate solution to transform both structured and unstructured data into actionable intelligence. "A big data platform can ingest a wide variety of disparate data structures at scale, run complex matching logic, execute predictive analytics, and process machine learning algorithms in real time. This should be the foundation to drive care management decisions through quantitative insights," he says.

Delaney concurs: "If you look at an integrative delivery network with 10 hospitals, they can be on a variety of different EMRs. When you begin to do an analysis with the data, you need to normalize them in a specific framework. It's a big challenge. There is a certain amount of effort required. On an industry level, one of the opportunities is what the master data are and what those elements are. This is what other sectors that are more mature have done—master data management. As health care matures, it will begin to do that as well, but it is still evolving."

Krithika Srivats, senior director of clinical health at Hinduja Global Solution Inc, offers a slightly different view. "Big data are important, but the focused approach on all the different components of the analysis of that big data gives only a 2-D perspective. We also need to understand the subtle nuances of that information," she says.

How Staff React to Data
As new technologies emerge, the concept of reacting to data in real time becomes more plausible. However, to make it part of health care professionals' daily routines will take time.

"At the highest level, the reality is that organizational transformation is a blend of people, processes, and technology," Delaney says. "Dramatic acceleration and simplification really provide the technology solution that can do now what couldn't be done before, shifting analysis of large amounts of data into real time. By far, we're better than we were five years ago. What remains, though, is the challenge of people and process. We need to incentivize people culturally and make it work in their workflow. We need the right people and culture, and also the right processes."

Placing this new treasure trove of valuable data with the appropriate parties also is key.

"Cool analytics doesn't get you anywhere unless used to move the needle. Get it in the hands of those who will use it to do something different clinically or operationally," Guest says. "You have to put information into the hands of care managers, providers, and physicians, but they will react to it in different ways. You have to present the information in a way that's usable to that stakeholder."

Foster believes ensuring staff members react properly to the information they receive goes back to one of the three Cs. "The culture has to be set by the organization," she explains. "The culture has to be that we recognize information is a strategic asset. But if there are hundreds of metrics in a strategic plan, then it's not a manageable scenario. We have to focus. We have to get good-quality data to inspire confidence, and then we need the right culture and competencies. Culture is the biggest paradigm shift in a value-based environment."

Foster advises organizations to take the concept of data as valuable assets to heart through the following mnemonic:

A — Align data and metrics against strategic imperatives.
S — Select the appropriate data source.
S — Store data where they can be aggregated, deidentified, deduplicated, and managed within privacy regulations.
E — Enrich the process by answering questions and converting responses into information.
T — Tell the story. What is the visualization strategy for the story? How do stakeholders consume the data and interact with them?

To implement a successful data analytics initiative, Currier says it is critical that workflow not be interrupted and any gaps be identified and addressed. "We don't want to take providers out of their workflow," he says. "We need to have IT play a key role in helping the doctor by seeing what it is he or she needs to do. That's the way to make the data we're talking about accessible to the user. How that workflow can engage the doctor and nurse and address the quality measures that are to be achieved."

— Susan Chapman is a Los Angeles-based freelance writer.