A Model of Health

Centerstone Research Institute mines patient data with IBM SPSS predictive modeling to optimize treatment and enhance clinical outcomes

Published on 25-Mar-2011

"Our goal is to deliver personalized medicine based on real-world experience. IBM SPSS predictive modeling is a key tool to help us achieve that." - Tom Doub, Chief Operating Officer, Centerstone Research Institute

Centerstone Research Institute


Deployment country:
United States

BA - Business Analytics, BA - Business Intelligence, Smarter Planet

Smarter Planet:
Smarter Healthcare


In the ongoing national debate about controlling medical costs without sacrificing quality, one strategy has stood out as holding tremendous promise: evidence-based medicine. The approach, which has garnered praise from physicians and politicians alike, seeks to apply the best available scientifically gathered evidence to medical decision-making in an effort to ensure the best clinical outcomes for patients.

Business need:
To improve patient outcomes and contain costs, Centerstone Research Institute (CRI) wanted to provide clinicians at mental health clinics with accurate guidelines around the efficacy of various treatments for mental health disorders.

CRI created a data mining solution using IBM SPSS Modeler that analyzes multivariable patient treatment information to provide individualized predictions of patient outcomes and help clinicians optimize treatments for mental health disorders.

By compiling and modeling data from patient experiences, clinical outcomes, and other sources, clinicians can now understand the efficacy of medical treatments within months instead of years, an approach that could potentially help contain the costs of healthcare. • Yielded 50% improvement in choosing the correct course of treatment during pilot programs

• Helped network of mental health clinics incorporate a “practice-based evidence” approach to treating patients • Supports clinical treatment model that leads to better treatment decisions, lower costs, and fewer problems related to side-effects • Helped clinicians see rapid results from treatment choices • Lowered overall operating costs, which will help the clinics weather fluctuations in federal funding and insurance reimbursements

Case Study

In the ongoing national debate about controlling medical costs without sacrificing quality, one strategy has stood out as holding tremendous promise: evidence-based medicine. The approach, which has garnered praise from physicians and politicians alike, seeks to apply the best available scientifically gathered evidence to medical decision-making in an effort to ensure the best clinical outcomes for patients.

Centerstone Research Institute (CRI) is on the forefront of evidence-based medicine. The private, non-profit company works with the community mental health centers of Centerstone to conduct clinically relevant research for the benefit of individuals with mental illness. Now, with the help of predictive analytics and IBM® SPSS® Modeler, researchers at CRI are taking evidence-based medicine a step further: They are using direct feedback from patients to steadily improve Centerstone’s knowledge base and help clinicians understand which treatments have the most likelihood of being successful. It’s an alternative approach called practice-based evidence.

“Practice-based evidence is actually deriving information from live clinical practice – from real-world populations,” says Tom Doub, Ph.D., chief operating officer for CRI. “Then you can apply that information back into the clinical practice and adapt it to fit the needs of the actual patient population and all the real-world variety that exists.” To analyze the data, CRI implemented IBM SPSS Modeler, a predictive analytics solution that reveals patterns and trends in data to better understand what factors influence outcomes. Researchers can use those patterns to build clinical decision support tools within the electronic health record that generate individualized treatment recommendations for future clients. With the intelligence generated from the data, CRI has been able to help clinicians significantly improve patient outcomes – the center’s No. 1 priority – and lower costs.

Doing more with less
Centerstone’s network of more than 130 nonprofit community mental health locations in Indiana and Tennessee serves more than 75,000 people each year with illnesses ranging from depression to drug addiction to stress-related disorders. CRI’s objective is to help guide the Centerstone clinics by conducting clinical research on a variety of frequently seen illnesses, providing clinicians with actionable information for its clinical and business management practices.

However, like healthcare providers and nonprofits throughout the country, Centerstone has been grappling with funding cuts, lower reimbursement
schedules, and increased demand for mental health services. “We realized we were going to have to figure out how to make better decisions with the limited resources we have,” Doub says.

CRI’s response was to design predictive models that could analyze the effectiveness of various medications and modalities to see which were the most beneficial. But instead of using a standard controlled trial approach – which can take years to complete, yield impractical solutions, or produce
findings that are outdated by the time they reach practitioners – the clinicians at Centerstone could use CRI’s patient-generated data as a tool to predict outcomes.

“One of the biggest problems we’re seeing is that there’s just too much information for one person to be able to make the right choice,” says Casey Bennett, lead data architect at CRI. “Take depression. At the moment, there are 20 different possible medications to choose from and almost no way to know which one will work. So clinicians are forced to make an educated guess. If it works, great; if not, try another one. But that’s expensive and not really the best way to proceed.”

Personalized medicine
CRI also didn’t want to base its decisions on data gathered on patients who bore little economic or demographic resemblance to Centerstone’s
Tennessee and Indiana clients. “A lot of people are doing what they call clinical decision support, but it’s often based on embedded, hard-coded
rules from 10 years ago,” Bennett says. “But they’re not personalized to individuals. They’re based on what works for 60 percent of people, based on average populations. That’s not the model we wanted.”

To build more accurate models, Centerstone configured Modeler to analyze 14 variables across more than 9,000 patients, including socioeconomic
status, demographic information, and a range of diagnostic and clinical data. Researchers also incorporate multiple sets of outcome measures from internal electronic health records as well as a statewide outcomes collection system. The end result is a set of predictions about the effectiveness of various treatment options for each patient based on his or her unique characteristics.

One of the most valuable inputs to the model – and one that CRI is just starting to use – is direct feedback from the patient, which researchers gather at every session. “We want to know what the patient’s perception of the treatment is,” Bennett says, “whether there is any improvement from their point of view. So we go directly to the patient and ask them how they’re doing, and how treatment is progressing.”

Each response is measured on 1-to-10 scale and collected throughout the course of treatment. The feedback helps predict the quality of the relationship between the medical provider and patient, which in turn becomes a good predictor of positive health outcomes. Over time, as data from more patients are fed into the model, and as the model’s algorithms “learn” from their predictive successes and mistakes, researchers will develop increasingly accurate recommendations for individuals – in essence, a kind of “artificial intelligence” approach. “Our goal is to deliver personalized medicine based on real-world experience,” Doub says. “IBM SPSS predictive modeling is a key tool to help us achieve that.”

Better clinical outcomes
Although CRI’s modeling project is still in the pilot stage, the practicebased evidence approach has shown every indication of succeeding. To date, intelligence derived from patient feedback has boosted the ability to choose the most appropriate treatment option for a given patient to between 70 and 75 percent.

“We know from health research literature that patients are only diagnosed correctly 50 percent of the time at first pass,” Bennett explains. “On top of that, patients are only given the correct treatment about 50 percent of the time. That means we’re only achieving correct diagnosis and treatment rates of about 25 percent at first pass. With IBM SPSS Modeler, we can increase that rate significantly and deliver more personalized medicine.”

And by eliminating less effective courses of treatment, Centerstone will be able to lower operating costs, an outcome that helps secure its own future along with its patients’. “Each day, we are improving clinical care by constantly re-evaluating data and applying those patterns to better understand future events,” Doub says. “Furthermore, by feeding the same information to management, it improves accountability. Finally, the transparency of the system empowers end-users and frees up resources that previously went to tracking and maintaining less efficient processes.”

About IBM Business Analytics
IBM Business Analytics software delivers complete, consistent and accurate information that decision-makers trust to improve business performance. A comprehensive portfolio of business intelligence, predictive analytics, financial performance and strategy management, and analytic applications provides clear, immediate and actionable insights into current performance and the ability to predict future outcomes.

Combined with rich industry solutions, proven practices and professional services, organisations of every size can drive the highest productivity, confidently automate decisions, and deliver better results.

Products and services used

IBM products and services that were used in this case study.

SPSS Statistics Professional, SPSS Modeler

Legal Information

© Copyright IBM Corporation 2010 IBM Corporation Route 100 Somers, NY 10589 US Government Users Restricted Rights - Use, duplication of disclosure restricted by GSA ADP Schedule Contract with IBM Corp. Produced in the United States of America October 2010 All Rights Reserved IBM, the IBM logo, ibm.com, WebSphere, InfoSphere and Cognos are trademarks or registered trademarks of International Business Machines Corporation in the United States, other countries, or both. If these and other IBM trademarked terms are marked on their first occurrence in this information with a trademark symbol (® or TM), these symbols indicate U.S. registered or common law trademarks owned by IBM at the time this information was published. Such trademarks may also be registered or common law trademarks in other countries. A current list of IBM trademarks is available on the web at “Copyright and trademark information” at www.ibm.com/legal/copytrade.shtml. SPSS is a trademark of SPSS, Inc., an IBM Company, registered in many jurisdictions worldwide. Other company, product or service names may be trademarks or service marks of others.