Big Data Analytics Use Cases for Healthcare IT

Though many healthcare organizations still regard analytics as a “rather leisurely activity” focused more on data and less on actual analytics, the International Institute for Analytics sees healthcare “speeding up dramatically” in its adoption of analytics. Current technology makes it possible to create data, products and services — think analytics as a service — to help guide institutional decision-making.

A recent IIA webinar outlined the state of analytics opportunities for healthcare and presented five case studies of big data in action. These five scenarios, plus five others, highlight what’s possible.

More: Can Healthcare Big Data Reality Live Up to Its Promise?

As IIA sees it, analytics 1.0 largely involves descriptive analytics that comes from small sets of internal, structured data, Research Director Thomas Davenport says. Resulting reports tend to stay within IT departments, away from decision makers, and look back, not ahead.

Analytics 2.0 considers complex, large and unstructured data sets and develops products (not reports) that make information readily accessible. This represents the heart and soul of big data startup activity, he says, but remains largely confined to Silicon Valley.

Finally, analytics 3.0 bridges traditional analytics and big data, using “rapid, agile insight delivery” to put analytics tools at the point of decision. Examples include LinkedIn’s “People You May Know” and “Jobs You May Be Interested In” features, Davenport says.

IIA offers a five-part “prescription” for analytics 3.0:

— Use existing data management and analytics capabilities — Incorporate unstructured, large-volume data (such as doctor’s notes) and product/service innovation — Employ Hadoop and NoSQL as needed — Embed analytics into organizational processes and systems, so it’s available at the point of care — Make a doctor, pharmacist or other employee with medical credentials your Chief Analytics Officer

This will help healthcare organizations both establish and measure “analytics maturity,” Davenport says, which in turn demonstrates to executives and board members the potential of analytics.

The following slides show how 10 payers, providers and pharmaceutical companies are putting big data analytics to use.

Patients filed more than 1.5 billion prescriptions annually with Express Scripts, a pharmacy benefit management organization. The company is using data about these transactions to drive both behavior change and process improvement, IIA notes. Patients may receive customized messaging about less expensive ways to refill prescriptions, for example. Meanwhile, Express Scripts’ predictive analytics can identify patients at the greatest risk of skipping or otherwise missing doses and then proactively intervene.

More: Big Data Project Cuts Prescription Costs

Intermountain Healthcare, a Utah-based system of 22 hospitals, 185 health groups and an affiliated insurer, has partnered with Deloitte on two tools that use a provider’s vast collection of electronic health records (EHRs) — more than 90 million in Intermountain’s case — to perform health outcomes analysis. OutcomesMiner, announced in July, helps organizations better understand how certain factors contribute to patients’ outcomes; this helps researchers see how empirical evidence supports or refutes a hypothesis. PopulationMiner, meanwhile, helps users study the relationship between treatments and outcomes, with the aim of developing new medications or improving existing ones.

Insurer UnitedHealthcare has found several ways to use analytics, IIA points out. The firm studies social networks to detect potential cases of medical fraud and identity theft. It looks at speech-to-text call center data to find likely attrition candidates — who sound different than happy customers — and propose remedies. It tries to predict the likelihood that certain disease management programs will succeed, since patients respond to treatment plans differently. Finally, UnitedHealthcare and OptumInsights are investigating potential revenue sources from these analytics processes.

Related: Big Data Analytics Gold for the Call Center

The Boston-based Partners HealthCare system — which brings together Brigham and Women’s Hospital, Massachusetts General Hospital and Harvard Medical School, among others — is connecting its financial, operational and clinical analytics systems. To aid this effort, Partners has developed the Queriable Patient Inference Dossier, or QPID, which it describes as an “intelligence system for the EHR” that enables real-time queries, analytics and reports at the point of care (and in easy-to-read reports for administrative or executive staff). Roughly 5,000 physicians and other medical professionals currently use QPID, Partners says.

Analysis: How Big Data Will Save Your Life

Like UnitedHealthcare, insurer Cigna is also analyzing call center speech-to-text data in an effort to improve customer service. Beyond that, though, Cigna conducts analysis in what the IIA’s Davenport calls an “omni-channel environment” by looking at Web data and information from other external sources. The insurer is also testing, and then analyzing, disease management interventions that occur over the phone.

Feature: Healthcare Industry Sees Big Data As More Than a Bandage

At the recent Strata Rx conference, the Children’s Hospital of Philadelphia demonstrated its Center for Biomedical Informatics (CBMi) Care Assistant. Built using the JBoss Drools Rules Engine, the Care Assistant is a clinical decision support framework that uses 1,500 rules to search 200 variables to return recommendations to the hospital’s EHR users, CBMi Senior Programmer Jeremy Miller says. In an 8-month trial involving more than 1,500 premature newborns, clinicians using an advanced Care Assistant version were more likely to use the correct growth chart and development milestones, administer an RSV vaccine and recommend a blood pressure screening before the baby’s first birthday. Tellingly, those using the rules engine wouldn’t give it up after the trial, Miller says.

One percent of patients are responsible for roughly 20 percent of all U.S. healthcare costs. These patients often have multiple chronic conditions, making cost-cutting tough. The “next 5 percent,” though, can present significant savings opportunities. A joint analysis of 30 million commercial claims by McKinsey and analytics vendor BeyondCore identified numerous microsegments — patients with heart trouble and fractured hips, or those with diabetes or septicemia — that can cost a hospital system as much as $100,000 per patient. A 500-bed hospital may only see 19 such patients per year, notes Tim Darling, who leads R&D within McKinsey’s Objective Health arm; a single, full-time care manager arranging home visits and helping patients keep appointments and take medications can deliver big savings.

Medicare reimbursement for acute cardiology favors continuous care over episodic, rush-to-the-hospital care. But less than half of patients hospitalized following a cardiac episode complete semiweekly rehabilitation for 18 weeks, as directed. Behavioral economics suggests that simply telling patients to participate won’t work, says Lise Worthen-Chaudhari, research assistant professor at The Ohio State University Wexner Medical Center. To improve patient engagement, OSU and analytics firm farsite studied what makes patients skip appointments; factors include co-morbidities, transportation and seasonality (dark winter afternoons demotivated patients). In a subsequent trial modeled after text4baby, patients’ families would send encouraging text messages over the 18-week rehabilitation period. So-called “hotspotter” friends — deemed by analysis to be positive influencers — also sent texts.

The doctor’s appointment is a good time to collect patient data, says Hulya Emir-Farinas, senior principal data scientist at Pivotal. But most patients go months between appointments, patient self-reporting is largely unreliable and both EHRs and insurance claims are often incomplete. Data scientists aiming to examine population health, then, should look to public health sources such as the FDA Adverse Event Report System. This, too is incomplete; it also contains duplicate events (reported by multiple entities), and it needs to be standardized and scrubbed (lawyers submit disproportionately large numbers of claims). Do the legwork, though, and there’s big data gold to be found.

More: 6 Big Data Analytics Use Cases for Healthcare IT

The all-payer claims database unites public health data (namely, birth and death records) with service dates, procedural codes, specialists seen and charges, all gleaned from claims data. The challenge: obtaining data from multiple institutions, since patients visit more than one healthcare provider. Beginning in 2002, the Camden (N.J.) Coalition of Healthcare Providers has collected data from the city’s three major hospitals, encrypted and standardized it, matched it to GIS data and, finally, identified “high utilizers” of healthcare services in Camden. Critically, linking data sets helped researchers identify 961 high utilizers across the three hospitals — nearly 33 percent of the total number of patients — who would’ve otherwise been missed if each hospital’s data remained siloed.

A recent survey by the Society of Actuaries, who polled more than 250 executives at U.S. hospitals, health systems and insurance companies, shows that most view population health management as the type of analytics initiative that will have the biggest short-term impact — that is, within the next two years. Over the next decade, though, population health management drops in importance for payers, while clinical decision support and chronic disease management rise. Both payers and providers deem the task of reducing hospital readmissions more important in the short term than the long term — which suggests that big data in healthcare, by 2023, may be poised to solve larger, more complex issues not mandated by healthcare reform.