HMT Newsletter Sign Up

 

 

 

 

 

 

 

 

 


 

 

 

 

 

 

 

 

 Analytics

Emerging directions in analytics

Predictive analytics will play an indispensable role in healthcare transformation and reform.

Email this article to a friend
  

   By Peter Edelstein, M.D., January 2013


The era of predictive analytics has arrived. Stage II and III meaningful use and the ongoing surge of big data demand predictive models and analytics that enable analysis and interpretation of large volumes of increasingly complex data. Healthcare organizations (HCOs) must tap data-mining technologies with the capacity to analyze data from varied perspectives and translate outcomes into meaningful information. By doing so, HCOs will be better positioned to make timely, evidence-based decisions, improve outcomes, streamline operations, boost cost efficiency, minimize risks and increase revenues.

This year promises to be an exciting one for predictive analytics, with major achievements and milestones in the prevention of early post-discharge readmissions, fraud detection, patient engagement and provider investment and partnership. Among the emerging trends for predictive analytics in 2013 are:

Predictive analytics will target identification of early readmissions risk. Thanks to the Hospital Readmissions Reduction Program, which reduces reimbursement to hospitals with excess readmissions, more than 2,000 hospitals that demonstrated above-average 30-day post-discharge patient readmission rates were recently penalized up to 1 percent of their Medicare reimbursement in fiscal year 2012, resulting in total penalties of about $280 million.

HCOs are spending significant money and volumes of resources in a desperate attempt to reduce early readmissions, much of which is due to factors out of the control of the hospitals (homelessness, transportation challenges, unemployment, drug or alcohol dependency, etc.). While predictive analytics cannot solve these core patient challenges, by incorporating clinical, demographic and socioeconomic data, and through the use of regression analyses of large-scale databases to generate weighted drivers of risk, predictive modeling can much more clearly identify which inpatients are at the highest (or growing) risk of post-discharge readmission in real or near-real time. Clinicians can take action while a patient is still hospitalized, and simultaneously target the patient for post-discharge processes aimed at preventing early readmission. Such predictive analytics allow HCOs to not only reduce readmission rates (and the associated reimbursement penalties), but also to reduce costs and resources allocated for prevention programs.

The results of predictive analytics are impressive. HCOs, such as Intermountain Medical Center (Murray, Utah), have reduced readmissions by relying on predictive analytics. But HCOs will soon come to expect more from vendors, including additional rigorous predictive models on the risk of mortality, sepsis, transfers to an intensive care unit and excessive length of stay. Forward-looking vendors must fulfill these expectations.

Predictive analytics will facilitate population health management and value-based accountable care. Risk analysis is impossible unless HCOs derive meaningful analytics from data. The algorithms of predictive modeling can analyze hundreds of data points – retrospectively or prospectively – to predict risk for populations and individuals. Such “intelligent analytics” offer HCOs the information and predictions they need to implement focused processes and guide clinical decision making.

Analytics will be invaluable as HCOs tackle the surge in Medicaid beneficiaries generated by the Patient Protection and Affordable Care Act. Medicaid is already expensive, with costs reaching $466 billion in 2011 for approximately 60 million beneficiaries, and just 5 percent (2.4 million) of beneficiaries accounting for half of total Medicaid spending.

HCOs face a unique set of challenges in caring for newly eligible patients, such as minimizing or avoiding costs, working within the confines of bundled payment or shared savings models and tackling problems prevalent within low-income Medicaid populations (such as chronic disease, high emergency department usage, dual eligibility, dual medical and behavioral diagnoses and prevention of admissions and readmissions).

HCOs will increase investments and rely upon predictive analytics. HCO adoption of health data analytics will increase over the next five years, according to a 2012 Frost & Sullivan report, with half of hospitals relying on analytics by 2016. A 2012 Black Book Rankings survey provided a similar insight: 84 percent of HCOs without a clinical decision-support (CDS) system still expect to implement at least one new or added clinical analytics tool in 2013.

Predictive analytics is far from reaching widespread acceptance. A 2012 survey by the eHealth Initiative (EHI) and the College for Health Information Management Executives (CHIME) revealed that while more than half of respondents reported using ad-hoc queries, data mining and data warehousing, less than half reported use of exploratory data analysis and online analytical processing. Only 23.6 percent of respondents reported use of predictive modeling, while 58.3 percent said that they focused resources on retrospective analysis.

Still, there is reason for optimism. According to a 2012 eHealth Initiative report, HCOs will be on the lookout for solutions that deliver standardized data across systems, provide infrastructure to support analytics, safeguard data privacy and security, and deliver practical, usable results. Again, vendors must deliver on the clinical, financial and operational expectations of clinicians and executives.

Health systems and academic medical centers will likely be the first to step up to the plate in using predictive analytics. For example, the University of Pittsburgh Health System has already announced its intention to invest more than $100 million in data analytics, including creation of a data warehouse with clinical, financial, operational and genomic data from more than 200 affiliated facilities. Among the goals: personalized medicine, population health management, increased operational efficiency and more accurate predictions of patient risk and treatment effectiveness.

In the future, predictive analytics innovation will likely be driven by creative partnerships, such as the five-year collaboration between the Regenstrief Institute and Merck that uses clinical data to design interventions for chronic conditions, such as diabetes, heart disease and osteoporosis.

Predictive analytics will emerge as a core strategy for detecting fraud. The Centers for Medicare and Medicaid Services (CMS) has already committed to a $90 million system that relies on predictive analytics to battle Medicare and Medicaid fraud. Focused on identification of high-risk claims, providers and analysis of fee-for-service data, the fraud prevention system probably will not meet its initial January 2013 implementation goal. However, recommendations from the Government Accountability Office (GAO) related to integration of fraud reporting and payment processing systems will likely keep the project on track. The Department of Health and Human Services’ Office of the Inspector General (OIG) also plans to use predictive analytics to detect Medicare billing abuse within electronic health records, according to the Center for Public Integrity.

Predictive analytics will facilitate patient engagement. Companies have developed systems that leverage national, local and patient encounter data to drive communication, education and behavior change. By identifying patients who are likely to develop a condition or need a medical procedure or service, or those who might develop a chronic disease, predictive analytics offers a solid foundation for communication and education campaigns and disease management programs. HCOs can choose all available means of persuasion – from mobile alerts and patient portals to telephone coaches and real-world classes – to deliver the right message to the right patient at the right time using the right medium or technology. Backed by the insights of predictive analytics, these engagement campaigns will help direct patients to appropriate providers or service locations, deliver ongoing education or disseminate targeted messages on disease and condition prevention, testing and monitoring.

Looking forward, predictive analytics will play an indispensable role in healthcare transformation and reform, generating positive results in patient outcomes and cost and resource utilization efficiency, prevention of early hospital readmissions, patient engagement and fraud reporting and detection. HMT

About the Author

Peter Edelstein, M.D., is chief medical officer, Elsevier/MEDai. For more on Elsevier/MEDai: www.rsleads.com/301ht-201


Tags:  Analytics