How Healthcare Data Helps Caregivers 2

Dramatic change is coming to healthcare--to patient treatment, providers' business models, and the government's role in regulation and reimbursement. There are many drivers, and not everyone's on board, but the one defining factor is that these changes all depend on the intelligent use of data.

Large healthcare organizations already use business intelligence and analytics tools as a standard part of their financial and administration processes--to streamline billing, manage financial performance, allocate staff and equipment, better manage patients as they move through the organization, and uncover revenue opportunities. They're also starting to use these tools to improve their insight into the effectiveness of patient treatments, and to move toward the quality healthcare objectives and pay-for-performance metrics that are becoming an integral part of federal funding and reimbursement.

Healthcare industry BI was a $600 million market in 2009, and it will grow faster than any other BI vertical industry in the next five years, says IDC analyst Dan Vesset. Increased focus on financial performance management, labor productivity, cost control, and analysis of billing, payments, bed occupancy rates, and patient treatment will drive that growth, Vesset says. In particular, he foresees 10% annual growth in applications that use BI and analytics to manage patient interactions.

In the future, BI and analytics will have more impact on clinical decision making. Healthcare providers will be able to analyze treatment outcomes based on who provided them, what treatment options were chosen, and where they were given, among other factors. They'll be able to look at what they spent on personnel and other resources and see how those decisions affected patient outcomes.

As the experience-based insights that BI and analytics provide expand and mature, healthcare will enter a new phase of personalized medicine, where treatments are carefully tuned to individuals' needs based on extensive analysis of personal genomics and aggregate data about disease risk in select populations based on analysis of genetic markers.

What it Takes To Build A BI Infrastructure

Large healthcare organizations are leading the way. They have the budgets for the considerable up-front investment in data warehouses, data marts, and data integration tools needed to build the information management infrastructure for BI and analytics.

St. Joseph Health System, a $4.3 billion not-for-profit provider, uses Microsoft's Amalga Unified Intelligence System as a common BI data store. It loads patient data from its Allscripts, GE, and Meditech electronic health record (EHR) systems along with financial, staffing, and other data into Amalga, so that users can do reporting and analysis against this data store rather than against the primary sources.

The Amalga database contains about 43 TB of structured and unstructured data, generated by treatments received by 2.3 million patients, says Larry Stofko, CIO of St. Joseph, which operates 14 acute-care hospitals as well as home health agencies, outpatient services, and physician groups in California, New Mexico, and Texas. Before information goes into the Amalga system, St. Joseph uses hosted data integration middleware to cleanse, merge, match, and otherwise prepare patient records and data for BI reporting and analysis.

St. Joseph's nursing and other managers use the system to track trends that might indicate a need to change how they assign caregivers, supplies, or budgets. Its finance department uses the system to analyze trends, including how product lines are performing and at what capacity. Physicians use it over the Web to access their patients' hospital information, lab results, and other reports.

St. Joseph has linked Amalga to its Microsoft HealthVault per- sonal health record system. When patients opt into Health- Vault, Amalga passes them electronic copies of documents such as clinical summaries of hospital stays and discharge instructions. The ability to share those types of records lets St. Joseph increase its engagement with patients, Stofko says--patient engagement is part of the federal requirements to demonstrate meaningful use of EHRs.

As more information is pulled into Amalga, St. Joseph plans to increase reporting and data mining so it can correlate what physicians and clinicians are doing for patients with outcomes and costs of treatment. "We want to use Amalga for reporting and analysis in real time--or at least with a more recent historical perspective--on our actual experience with our patients," Stofko says.

Most physicians in standalone practices don't use BI applications, as most of those require more IT support than small medical groups have, and many practices are still transitioning to EHRs. However, as these practices become part of provider networks and health information exchanges, BI and information management tools will become more prevalent.

One example is the recent launch in Minnesota of the first statewide electronic-ordering and decision-support system for diagnostic testing. The system, available to 9,000 physicians, employs Nuance Communications tools to help doctors pick the most appropriate diagnostic imaging tests. BI software then correlates physicians' e-ordering trends with patients' clinical outcomes, providing insight into the value of these tests (see story, p. 14).

The Silo Problem

For BI and analytics to take off in clinical settings, a lot depends on improving healthcare organizations' data, providing better integration, and increasing access to more data sources, both structured and unstructured.

Most hospitals and provider groups are run more like agglomerations of independent businesses than as cohesive entities. Cardiology does things one way, obstetrics another. This setup makes it challenging for IT to help managers use information effectively.

The problem is particularly acute in hospital purchasing processes, where physicians often drive the selection of the devices, equipment, and drugs needed for treatment, and organizations fear that if they try to take that authority away, medical talent will move to competing providers. Taking advantage of the lack of coordination, drug and other suppliers are able to offer different departments and practitioners within the same hospital different pricing for the same goods.

Change is coming, however, as Medicaid and Medicare reimbursement policies start putting more emphasis on efficiency and clarity in purchasing. Some organizations are already using BI to gain insight into costs across their hospitals.

Mount Sinai Medical Center, a 1,200-bed New York City teaching facility that treats nearly 60,000 inpatients and more than 500,000 outpatients a year, is using Information Builders' WebFocus to help department managers monitor expenses. Top executives, business managers, physicians, and administrators now see the same reports, with security filters ensuring that each user sees only appropriate data, says Gad Malamed, Mount Sinai's director of IT. Dashboards let physicians and administrators monitor what's going on, as opposed to what they previously had to do: "Wait months for reports long after the fact," Malamed says.

WebFocus also lets Mount Sinai track and analyze how expenses and payments for each of its physician groups compare with industry benchmarks. The medical center can then have fact-based discussions with its doctors to determine whether treatments that are more expensive than industry norms are actually delivering better outcomes.

Integrating Data Sources

To support BI and analytics, large healthcare providers are building data marts and data warehouses, turning to the likes of IBM, Informatica, Microsoft, Oracle, and SAS. Other vendors, including Healthvision, MobileMD, and Novo Innovations, provide information integration tools that address specific requirements for EHR systems and health information exchanges.

Clalit Health Services, Israel's largest health maintenance organization, is a leading-edge example of how a large healthcare provider is knitting together information from many sources. Clalit integrates data from its 13 hospitals and 1,000 clinics into a 3-TB enterprise data warehouse, based on Microsoft SQL Server. Patient clinical and financial data is fed into the system from clinical applications and ERP systems. The warehouse contains 10 years of data and supports analysis of clinical histories for millions of patients. "Complex ad hoc queries that access 10 GB of data are very common," says Mazal Tuchler, Clalit's BI division manager.

Clalit has a secondary data warehouse, containing 1.5 TB of financial and patient care data, dedicated mainly to performance management. SAP Business Warehouse and Business Objects manage the secondary warehouse, which supports more than 250 OLAP cubes for 5,000 managers, assembling views of different slices of performance data.

Clalit performs data mining to analyze the accuracy of its forecast models. It's now able to develop models that better allocate resources, improve operational efficiency, and optimize healthcare delivery.

Consistency And Quality

While data integration is important, data consistency and quality is critical to BI and analytics. The decentralized operational data silos so common in healthcare organizations frequently contain inconsistent or incorrect data about patients, treatments, physicians, and suppliers. Data definitions and other metadata often aren't consistent.

Data quality and consistency problems carry obvious risks to patient care. They also contribute to higher costs as errors proliferate.

A range of data-quality tools address these problems, from vendors such as DataFlux, Harte-Hanks Trillium Software, IBM, Informatica, Pitney Bowes, and SAP Business Objects. These tools usually are first applied to name and address data, but in healthcare, codes and other identifiers for testing, billing, and other functions must be addressed as well. Most tools use rules engines to determine whether data meets certain standards. Many employ logic to discover matches and links between pieces of data, such as between two members of the same household. Data quality systems also come with tools that monitor sources so that IT can maintain quality and consistency.

Some healthcare providers are implementing higher-level information integration so that data-quality problems are remedied across the enterprise. To do this, they must make sure that all data associated with a particular patient is linked, so that it can be updated in multiple systems without inadvertently exposing the data to security or privacy holes. This comprehensive approach to defining and linking data objects is critical to anti-fraud efforts healthcare organizations must have in place.

Some healthcare providers are implementing master data management--practices and tools that focus on achieving information consistency and quality throughout an organization. MDM practices typically involve using data quality, profiling, matching, and other tools. One common approach is to establish either an index or repository that contains a "gold" reference copy of information about a patient that can be applied as data from multiple sources is accessed. Reference data is particularly useful in healthcare for applying privacy and security rules governing access to patient data.

MDM is a relatively new and immature concept but will be important in improving BI and analytics in healthcare for two main reasons.

First, it improves data quality and consistency earlier, before data arrives in data warehouses, relieving pressure on processes used to prepare data for BI and analytics. Those "transformation" processes can cause data quality and consistency problems to proliferate, such as when different spellings of a patient's name are loaded into the data warehouse from multiple systems.

Second, medical providers often need to see all relevant data about a single patient. This single view is important not just for patient care, but also to manage security, privacy, and fraud detection. MDM tools for data matching and profiling can help IT locate all related data about particular entities that may be fragmented across multiple systems.

Two providers in the St. Joseph Health System--St. Jude Medical Center and St. Jude Heritage Medical Group, both in Fullerton, Calif.--use a hosted health information exchange run by Accenx Exchange. Accenx implements the IBM Initiate Master Data Service, providing a master patient index that uses the Initiate algorithms to examine data flowing through the HIE from St. Joseph's primary data sources, such as labs, specialists, and operations units. The algorithms confirm the quality and consistency of patient data.

"We make sure that we have distinct matching of our patient data so that we don't have duplicate questions upon registration," says St. Joseph CIO Stofko. "As patients go from hospital-based services to physician offices, radiology labs, and other departments, we don't want them to have to register more than once." St. Joseph plans to roll out more enterprise master patient index systems in its next fiscal year.

Dashboards: Making BI Actionable

Dashboards are the preferred choice for displaying BI and performance management information in an easy-to-understand way. When integrated with underlying data integration and quality technology, they provide the user interface for an enterprise BI architecture.

Dashboards for executives and managers responsible for financial performance display forecasts and actual data for each hospital and clinic in the provider's network. These users drill down into graphs of the actual versus forecast performance to gain deeper insights into each hospital's performance over time. Patient care dashboards display charts for particular events, such as patient contacts, referrals, and prescriptions.

Barnes-Jewish Hospital, one of the largest healthcare providers in Missouri, has deployed Tableau Software BI tools to ensure that staffing needs are addressed quickly. Barnes-Jewish has to keep track of more than 2,400 nurses and other caregivers who attend an average of 840 patients a day.

At the center of the system is a dashboard that integrates real-time data from eight sources and provides visual triggers to let managers know when they're in danger of being understaffed or short of needed expertise.

Barnes-Jewish's dashboard offers graphical, point-and-click interfaces that let users query the BI system and tap into data on how many nurses and other personnel are on call at a given time and where they're working. The system compares this information with data about how many patients are being admitted and what their care needs are.

Since deploying the system, Barnes-Jewish Hospital has reduced both over- and understaffing by a third, says Dr. Linh Dye, patient care services special projects manager. "Before, we didn't have information to react to," Dye says. "Now we can take action right away."

Barnes-Jewish's staff scheduling dashboard lets managers easily solve over- and understaffing situatons.

Better BI Equals Better Care

The growth of BI and analytics in healthcare will be fueled by strong demand for performance data and ultimately the desire for deeper insight into relationships between treatments and outcomes. IT must meet these demands not just with technology, but also with a better understanding of the cultural and political issues involved in getting nontechnical users to apply the fact-based insights coming out of these tools.

IT managers should look for opportunities to demonstrate the short-term value of BI. In addition, IT should develop a plan for scaling the numbers of users, amounts of data, and update requirements. And above all, remember that BI is only as good as the quality of the data you provide. (For more steps to BI success, see KEYS TO BI Success, below)

Keep in mind that these systems thrive on expanding access to data. And as long as the data is good, BI will give healthcare providers a critical tool for achieving the ultimate goal: quality healthcare at an affordable cost.

KEYS TO BI Success

It takes patience to get business intelligence and analytics right. User requirements often become clear only when users try out the software. And it can be a never-ending quest to open up access to the right data and to make sure that its quality is acceptable. Here are five factors to consider in selecting and implementing BI and analytics software:

>> Know your culture. BI and analytics aren't just technology. The ultimate goal is to improve how people make decisions and use information to take action. Are physicians, nurses, and other users open to having their decisions challenged by data analysis? Make sure users understand BI's role in their decision making. Look for venders that have a track record in healthcare and understand the culture.

>> Think big, act small. While it's important to see the big picture, aim initial BI projects at addressing well-defined needs. Look for quick wins that improve labor productivity or reimbursement processes. Choose BI tools that don't require substantial IT involvement and offer self-service and online services for users to create reports and query data.

>> Deliver substance, not just flash. Dashboards are the preferred interface to most BI and analytics. However, just because the Web interface is more visual than spreadsheets doesn't mean that it's easier for nontechnical users. Fit the dashboard to the user's role. Look for BI tools that you can customize to users' needs.

>> Determine how often to update data. How often do physicians, nurses, and other front-line users need their data refreshed? What about staffing managers? It's important to answer these questions as you gather user requirements and evaluate tools and services. Real-time or intra-day updating can be expensive, so make sure you really need it. But if you do, make sure your BI vendor can support it.

>> Don't neglect data quality. The worst thing that can happen to BI is for users to lose confidence in the quality of the data. Consistency and quality can be manageable problems in a single data source; but once BI systems integrate multiple sources, the challenges increase considerably. Map out the flow of data as it's used. Evaluate vendor tools with data quality and consistency in mind.