Matching the practices of business intelligence (BI) to health care institutions is a process that has been steadily — if unevenly — taking place over the last two decades. Like any other industry, health care adheres to the topics and buzzwords that touch any BI undertaking which include data warehousing, data quality, data integration, metadata management, governance and analytics.

Unlike other industries, the clinical side of health care is based on interpretations of care practices that change over time. And while health care institutions compete for customers, there is no one set of product or service standards — as would come with a new refrigerator — because the patient is in fact the “product” shared among many providers.

“We generate and use data like any other industry, but health care does not lend itself to the use of discrete data because the outcomes are necessarily fuzzy and ongoing,” says Dick Gibson, M.D., the CIO at Legacy Health, an integrated network of six hospitals with research and other facilities in Portland, Oregon. “Airlines have seats, schedules and know if you landed on time. In health care, we know you’re alive but the big money goes to broad sets of descriptive terms around patient care that are very qualitative.” For BI and performance management to have meaning, these terms need to be unified and reconciled in their definition and accuracy.

The challenges of data quality are multiplied by the number of codes and procedures managed in systems for lab results, pharmacy, check-in and other processes. Even within a single institution, the lack of standards, or ironically, the fact there are far too many, creates huge data quality and integration challenges.

But it’s a challenge providers must meet, both in response to consumer demand and regulatory interest in keeping down costs. That’s why information systems are increasingly required to bring a compliant and cohesive view to inpatient and outpatient facilities, right down to the bedside of the patient at the point of care. The approaches to meeting this challenge vary by hospital, with some opting for a single-vendor approach while others turn to multiple software partners. The evolution of provider software companies offers an interesting comparison to other industries, which have adopted enterprise resource planning tools only to find the tools lacking. Nonetheless, the data warehouse in the clinical setting is here to stay.

Two Views of BI

Business intelligence can be described as a process that leads to better decision-making. In its original and ongoing mission, a business analyst selects and extracts historical data from one or many databases. The analyst then structures and loads the data into a single repository (a data mart or data warehouse). Analytical and visual tools are then applied to perform trending and comparative studies that, sliced in different ways, create reports to measure performance and uncover opportunities for improvement.

In contemporary form, BI also takes an operational approach that gathers near real-time data to support ongoing processes. These include sales, marketing and customer interactions. Operational BI is more process oriented than the data warehouse model and is associated with key performance indicators, dashboards and scorecards that support performance management.

In the clinical world, health care providers have tuned operational workflows to fit processes from admission and treatment to checkout with a flexibility other industries might envy. A visitor to an emergency room or someone admitted for an inpatient or outpatient treatment might well regard the event as orderly and procedural.

Whether or not this appearance is true, the actual management and reconciliation of data that flows through admissions, doctor notes, labs and pharmacies becomes a huge challenge to implementing BI. And data complexity provides only a partial explanation of why the industry is a relative late-comer to the BI strategies employed in other economic sectors.

While current budgets are tight everywhere, the health care sector has traditionally been an IT spending laggard. Without accounting for scale, a 2009 Gartner Research note predicts that the financial services industry will spend more than six dollars on IT for every equivalent dollar spent by health care this year. But according to figures from HIMSS, the Healthcare Information and Management Systems Society, this under spending may change. Spending, which now sits at approximately 2 percent of total revenue, is expected to grow at a compounded rate of 7.5 percent through 2014. A 2008 snapshot of health institutions conducted by IDC subsidiary Health Industry Insights found that less than 20 percent had instituted an enterprise data warehouse. The study also found that more than 30 percent were planning to do so.

Setting Priorities

Despite the uneven spending on IT, health care as an industry aspires to the vision of a “digital hospital” with consolidated electronic records, but this does not describe the state of work under way. It’s a step-wise approach to the vision. Many institutions are still turning from paper to discrete software applications that might cover a single workflow, such as processing new patient admissions or facilitating orders for lab tests.

For CIOs, maintaining the vision of the digital enterprise requires tenacity. In 1999 when Indranil Ganguly arrived as CIO at CentraState Healthcare System, a 271-bed acute care and outpatient facility in New Jersey, clinical data was being captured electronically, but not in a way that supported decision-making. “When I and a couple of others joined the organization we saw technology that was simply as a surrogate for paper. It was not adding value, and maybe was impeding value by making it more onerous to retrieve information.”

Some of CentraState’s discrete applications offered light reporting and bits of clinical and BI value, but at a very low level.  What Ganguly chose was to rip and replace isolated or weakly connected applications with a more unified platform from Siemens Medical Solutions that helps integrate data from multiple operational processes.

CentraState is hardly alone in this regard: One of the most obvious trends in health care information systems is that hospitals and integrated care networks are widely replacing older applications and connections with newer platforms. An expectation is that new technologies will also take over much of the data quality and integration heavy lifting, in part to fuel BI reporting and performance management. At Legacy Health, Gibson has torn out an expensive platform that was still under construction and replaced it with software from Epic Systems Corp. The new rollout connects the emergency department, critical care, physician office billing, hospital billing, hospital and clinical operations across Legacy’s facilities.

From a financial perspective, as government and third-party payers look for ways to control costs, it’s incumbent on hospitals to churn more detailed information, says Marc Holland, a former IDC Health Insights analyst who now runs a practice called System Research Services. “Hospitals need to better understand their actual costs, reduce the cost of providing care and concentrate on increasing market share in services that are most profitable.” In order to do that, he says, providers need more than traditional diagnosis-related group, discharge summary and billing data. “You need information about the kind of care that was delivered to the patients throughout the course of their stay. Hospitals know they need a lot more clinical data in digital form than they’ve traditionally had.”

A Unified Transaction Record

The march toward clinical IT adoption — and the BI functions it would enable — long predates the current federal government incentives to adopt electronic health records. While many EHR programs today are executed solely to improve safety and manage clinical processes, they also aggregate data and enable connections to financial platforms.

“The rise of BI and clinical intelligence,” says Holland, “is due in part to the fact that many hospitals have several years of experience with full-blown EHR systems that are yielding some of the missing pieces of the puzzle in terms of the kind of care delivered, when it was delivered, who delivered it, how long the stay was and what the outcome looked like.”

The transactional focus of EHRs bears a canny resemblance to enterprise resource planning software, which sprang up across all industries in the 1990s through post Y2K as a means of centralizing transactional data. ERP evolved to support sales, finance, customer interactions, supply chain, human resources and other departmental transactions.

ERP has become a lasting standard and key contributor to BI. However, a lesson emerged when businesses found that transactional processing and data repositories by themselves did little or nothing to manage, measure or automate business processes. After Y2K, ERP vendors quietly dropped the initials. They wanted to be better known as platform vendors who acquired or created applications to manage the same customer interaction, supply chain or HR processes they were recording.

Health care CEOs and CFOs may recognize the connection because they already employ ERP systems on the financial side of their business. The strategies of a Siemens, Epic, Meditech or Cerner might well echo the post-Y2K messaging of SAP, Oracle, Baan, JD Edwards or PeopleSoft with good reason. Unlike industries with longer and deeper software development histories, health care is still dependent on a unique set of vendors with fewer supporting application partners to fill out the nuts and bolts of enabling BI for their specialized needs.

A mix and match strategy might apply at hospitals with a financial reporting track record, such as MemorialCare, a  four-hospital system in southern California. MemorialCare had already standardized financial and HR systems with PeopleSoft and was providing executive financial dashboards when it embarked on an EHR rollout in 2002, which is now about 80 percent complete. “We now have billing and integrated solutions from Epic on the patient side and integrated solutions from PeopleSoft on the financial side,” says CIO Scott Joslyn. “One of the things we’re working on now is to integrate core measures for payers, like the time it takes to administer an aspirin once a patient with heart attack symptoms arrives, back into the dashboards.”

Platforms and Standards

If the history of the ERP industry is any indication, EHR platforms present benefits and risks related to their relative maturity. Pre-integrated platforms lessen requirements for systems integration and help unify documentation for decision-support, BI and performance management. They help bridge workflows and processes formerly siloed in individual applications, and may come with uniform upgrades across application areas.

On the downside, platform investments cost many millions of dollars in a commitment that can lead to vendor lock-in. Product roadmaps don’t serve all interests and vendors offer products based on proprietary code that may or may not work well with newer technology practices. Some platforms are based on code written decades ago, which can make them more proprietary.

Customers also risk the effects of mergers and acquisitions in products that might be discontinued. For example, Denver Health is in the process of replacing a radio frequency identification system it installed to track wheelchairs, pumps and other hospital assets because the original vendor was acquired and the product was discontinued.

But Denver Health is also profiting from a Siemens standardization that extends across clinical and financial applications. “Seventy percent of our clinical systems and probably 90 percent of our financial systems are now on Siemens,” says Jeff

Pelot, chief technology officer at Denver Health. And Denver Health is upgrading to the latest version of the Siemens Suite, which comes with built-in analytics and workflow, which Pelot says will increase the ability to collect and leverage data.

Despite these plans, Denver Health must still turn to a variety of IT vendors to support clinical operations, picture archiving, lab systems, cardiology and outpatient pharmacy. Integration is accomplished with help from Siemens but mostly through an internal interface engine that allows data to be written once and read in many systems, including the data warehouse.

In the ERP world, platform vendors have grown more dominant and improved product lines steadily over the years. They have also acknowledged the need for specialized and heterogeneous technology infrastructures, a trend also proving true in health care.

“Under the very best of circumstances if you bought everything you could from the smallest number of vendors you would still have a substantial variety of information systems just because of the nature of health care,” says Vi Shaffer, an analyst with Gartner Research and 30-year veteran of the health care industry. “You’re always going to have interoperability and interface challenges; the standards bodies have done yeoman service over the years to help that along but we still have a long way ago.”

The risk of non-standardization also exists where inaccurate sharing of information might affect patient safety. An example arose years ago at Geisinger Health System where, in a pilot program, order entry and pharmacy data were mismatched and patients were sometimes ordered the wrong medications or doses. The error was corrected four years ago but a  BusinessWeek article recounted the episode as recently as April under the gloomy title, “The Dubious Promise of Digital Medicine,” and quoted James Walker, Geisinger’s chief health information officer, as saying that providers are thinking, “Look, let’s slow down.”

Walker now says his quote was taken out of context and that safety issues around data quality are an ongoing priority in every aspect of hospital operations. “We’re deadly in earnest about making electronic health records safer but there is no question in my mind that the EHR has already made care safer than it was before.” In fact, he says, “once you get used to having all that information in a form that’s usable, it’s scary to take care of patients without it.”

Standards also change outside the four walls of databases and institutions. An issue at Geisinger and other facilities is the transition to ICD-10, the latest global coding standard for diagnoses and disease established by the World Health Organization. “ICD-9 does not translate to IDC-10, and we need to understand how we’re going to make that work for billing, for our clinicians, for our data warehouse and other systems with comparable data,” Walker says.

An exception to the single-vendor approach is work under way at the University of Pittsburgh Medical Center, a sprawling $8 billion network of 20 hospitals and 400 outpatient sites. UPMC had tested a platform product and found resistance to functionality that poorly served different roles. Despite its monolithic structure, the platform was also creating new silos of data, so UPMC embarked on a strategy of data translation and integration between applications to provide “semantic interoperability” with a (partly owned) partner called dbMotion.

For example, Cerner software dominates inpatient systems in UMPC’s hospitals, while a different system from Epic is used by nearly 800 physician offices. “We’re mapping data from both those systems to common fields for medications, allergies, immunizations and problem lists,” says Dan Martich, chief medical information officer at UPMC. “We’re taking on documentation and mapping lab information from inpatient and outpatient systems and making sure we get all the encounter and demographic information correct.”

Work Ahead

Many of the most inspiring stories in health care come from dedicated data mining and analytic research programs that are addressing specific illnesses and unique patient care.

Yet, it’s clear that much of the work lies ahead. While providers may be anxious to claim government incentives for electronic records and share information with others, most are in stages of getting their own operations in order.

Many, like MemorialCare, haven’t yet tackled a data warehouse project because other priorities have to come first. This hasn’t stopped CIO Joslyn from taking creative approaches to pull financial metrics into a dashboard for executive reporting.

Others like Geisinger have built a data warehouse mostly tuned to improving clinical operations and safety but use a for-profit business unit to provide data mining results to partners.

Still others like Denver Health have built complex data warehouses to manage and measure an increasing amount of clinical and financial system data. “We’ve built a methodology for performance management, our data warehouse and gone hard after it,” says Pelot, the Denver Health CTO. “We’re always cleaning up data and figuring out the metrics that matter. But the base of it is you’ve got to have the information in there first so you can figure out what those metrics are, and we’re going to be working on that for a long time.”