Between electronic health record (EHR) systems, imaging systems, electronic prescribing software, healthcare claims, public health reports and the burgeoning market of wellness apps and mobile health devices, the healthcare industry is full of data that’s just waiting to be dissected.
This data analysis holds much promise for an industry desperately seeking ways to cut costs, improve efficiency and provide better care. There are victories to be had, to be sure, but getting data from disparate, often proprietary systems is an onerous process that, for some institutions, borders on impossible.
Data Is Data, No Matter the Source
Generally speaking, most healthcare organizations’ data comes from clinical, financial or operational applications. On its own, each type of data has a specific use, which The Institute for Health Technology Transformation (iHT2) outlines in a report, Analytics: The Nervous System of IT-Enabled Healthcare. Clinical data improves care quality and eases population health management; financial data helps hospitals conduct cost analyses pertaining to the bottom line, and operational data examines facilities management and resource utilization.
Put it all together, and organizations can start to assess larger issues such as staffing needs, efficiency and care quality. That’s why Laura Madsen, business intelligence (BI) evangelist and healthcare services lead at Lancet Software, sees no need to differentiate among the different types of data sources. “Data is data,” she says. “At the end of the day, it’s just bits and bytes…If we’re good data professionals, we should be integrating clinical data and business data.”
Government programs—and mandates—place added pressure on the healthcare industry while giving organizations reasons to take a good, hard look at analytics. The meaningful use program that offers financial incentives to use EHR systems, the accountable care organization (ACO) model of coordinated patient care, the concept of the patient-centered medical home and the increased emphasis on improving care quality all require a more sophisticated approach to healthcare data analytics.
Abundant Unstructured Health Data Makes Analysis Difficult
Of course, organizations can’t analyze data without first collecting it. In healthcare, the iHT2 notes, several factors complicate this. As much as 80 percent of healthcare data is unstructured, whether it’s in paper format or in free-form fields that need to be manually abstracted, and even the structured data—that which comes from the health information exchange (HIE) process, for example—is often inadequate for analysis. As a result, the report continues, providers end up using claims data from insurance companies to get a broad view of their own organizations.
When it comes to healthcare BI, size matters, says Madsen, who literally wrote the book on the topic. The country’s largest providers, namely Intermountain Healthcare and Kaiser Permanente, have been doing it for a long time, but the gap is “huge” for smaller providers. Most of these organizations see the value of BI, Madsen continues, but they can’t come up with a clear answer to the question, “What should we be doing?”
Most opt to focus on BI as it pertains to regulatory reporting needs. This makes sense, as hospitals must file upwards of 1,000 reports to government agencies annually. With such an apparent need, though, Madsen says it’s often difficult for organizations to take the next step and see how the data in those reports can be used to promote operational efficiency or other institutional improvements.
Luckily, the iHT2 report offers several suggestions. Assessing a patient population’s health needs, for example, can help organizations develop appropriate methods of service delivery while also identifying individual care gaps and even predicting which patients are likely to become seriously ill. In addition, evaluating provider performance can help drive quality improvement programs and also pinpoint reasons for variations in care.
It’s also worth noting what will not work. Under the Medicare Shared Savings Plan as well as the ACO model, the aim is to generate savings that ultimately lead to lower healthcare costs, so revenue cycle management tools won’t work, according to iHT2. In addition, today’s cost accounting systems are ill-equipped to measure the total cost of care, which needs to consider that an early hospital discharge saves money for one facility (the hospital) but represents a missed revenue opportunity for another (the long-term care facility). Finding the total cost of care, iHT2 says, requires a “sophisticated, episode-based accounting system for bundled payments.”
Healthcare Taking Data Analytics ‘Wins’ Wherever It Finds Them
Not all analytics systems in healthcare need to be sophisticated. At Springhill Memorial Hospital in Mobile, Ala., a recent automated medication dispensing cabinet system update came with Pandora Clinicals, an analytics package that has helped the facility reduce narcotics diversions.
The software, from Omnicell, tracks who removes narcotics from the medicine cabinet and when. Monthly reports help hospital management pinpoint outliers who dispense more medication than others. At worst, Clinical Pharmacist Joe Adkins says, this may mean a staff member is diverting the narcotics for sale or personal use—though it can also mean that a nurse or clinician has been proactive in treating a patient’s pain. The software doesn’t prove association, he says, but it’s the first lead and often helps staff spot discrepancies before they otherwise would.
Critically, getting data from Pandora Clinicals has little effect on overall workflow, Adkins adds. Reports are automatically emailed and use bar graphs as opposed to lengthy written records. In short, little data manipulation or math is necessary: “It’s a nice way to keep an eye on what’s going on without having to think about it.”
For payers, meanwhile, the aim is to improve the customers experience in a way that patients don’t have to think about it, says Bob Dutcher, vice president of marketing for predictive analytics firm InsightsOne.
In a 2012 pilot, which is now live, the firm worked with Independence Blue Cross (IBC) to help the Philadelphia-area insurer identify patients who were likely to experience customer satisfaction issues and provide outreach to nip those problems in the bud—sometimes three months before they’d otherwise arise, Dutcher says. (The company also helped IBC identify potential new customers as well as existing members who may benefit from services they weren’t using.)
To do this, IBC looks at data from its call center, to see which patients make frequent inquiries and therefore may need some extra attention. It also looks at data from member healthcare organizations, to see which medical procedures prompt the most follow-up inquiries from patients and also to see why an individual patient needed treatment. This analysis can alert IBC that a particular patient has “a high probability of a negative outcome,” which may trigger the insurer to send information about preventive care (to avoid repeat hospitalizations for the same condition) or long-term or home health services (if an upcoming procedure will have a long recovery period).
Such a proactive approach improves the overall patient experience, Dutcher says, while potentially saving healthcare providers money on unnecessary or repeat procedures. InsightOne calls this sort of analytics “predictive intelligence,” he says, and it lets analytics get specific enough to identify a “pattern of one” for a single patient.
Analytics Needs Talent, Data Warehouses; Both in Short Supply
Getting insurers to spend money on advanced analytics, as stated, is easier than getting healthcare providers to invest. But there are two key reasons providers can’t stay on the sidelines for long, says Cynthia Burghard, research director for accountable care IT strategies with IDC Health Insights.
One is the argument that a patient is more likely to participate in a wellness program (that back-end analytics has identified her as a good candidate for) if the recommendation comes from her physician as opposed to her health plan.
The other is that healthcare reform efforts of the 1990s failed largely because of a lack of data. “Not only was the available information limited to claims but it was retrospective and not in a format that was useful to physicians in understanding their current performance compared with targets,” Burghard notes in a recent report, Business Strategy: Analytics Leads Accountable Care Investment Priority. “Most discussions between payers and providers resulted in arguments about the accuracy and timeliness of the data.”
The emerging ACO model, introduced in healthcare reform as a way to shift the industry from a fee-for-service model to one centered instead on coordinated care, placed added emphasis on analytics and data warehousing technology. The need here is identifying patients who will benefit from a particular care program, engaging those patients in order to manage and improve their care and to incorporate such care interventions into a physician or clinician workflow, Burghard says.
Down the line, as the ACO model and coordinated care expand, organizations will increasingly see the need to examine unstructured data, sentiment analysis and other data sources—including, perhaps, predictive intelligence and the mix of clinical and business intelligence—in the context of the patient encounter and clinical decision support systems, she adds.
To the challenges presented by such advanced analytics, Burghard says healthcare providers will “need fairly sophisticated people to leverage [data] warehouses and take advantage of what [they] invested in,” and those who can afford neither a data warehouse nor the staff to manage them may find themselves pressured to consolidate or join larger integrated delivery networks.
This will disrupt the industry, no doubt. But in the end, having more data on hand—and being able to use it—will also improve the industry. “The adage ‘You cannot manage what you cannot measure’ applies to accountable care,” Burghard writes in her report. “In the 1990s, healthcare organizations lacked an understanding of the critical nature of patient compliance in the management of chronic diseases; the industry is better informed today and is investing in technology to share data among payers, physicians, and patients to improve outcomes.”
Brian Eastwood is a senior editor for CIO.com with more than 10 years of experience writing, editing and producing content for newspapers and the Web. He is primarily responsible for working with CIO.com's contributors and columnists, who cover topics such as cloud computing, big data, development and architecture, personal tech, the IT channel, business applications, BYOD, consumerization and business / project management. Brian's specific area of interest and expertise is healthcare IT. Prior to CIO.com, Brian was an editor at TechTarget and a newspaper reporter in the Boston suburbs. Outside the office, Brian is a history buff with a particular interest in postwar Europe and a runner who recently finished his 11th marathon.