Follow Datanami:
April 25, 2012

Taking the Shock out of Sepsis with Big Data Analytics

By Mark Adams, Ph.D., Reechik Chatterjee, M.A., Jeni Fan, M.S., Juergen Klenk, Ph.D., Yugal Sharma, Ph.D., [Booz Allen Hamilton] & Dr. Robert Taylor, M.D., Dr. Joseph Drozda, Dr. Timothy Smith, M.D. [Mercy Health]

Electronic health records (EHRs) are slowly beginning to permeate more widely and deeply into our health care system. As part of this transition, increasingly the traditional file rooms are being replaced with electronic repositories of data. While the benefits of this effort are widely discussed, much is yet to be realized.

Why? Because integrating and analyzing data from the existing EHRs has been a significant challenge, particularly because existing EHR systems often were primarily intended to facilitate the business and payment functions, rather than on improving quality of care.

In addition, the complexity of that data, and of the biomedical questions that needed to be asked, significantly complicate the process of analyzing and deriving useful knowledge from the growing sources of electronic health information.

Like other industries, healthcare now has a “big data” problem.

Booz Allen Hamilton wanted to understand how this rich data that is being compiled by hospitals and healthcare providers could become an asset for patient care – leveraging advanced data collection and analysis to reduce patient risks and improve care.

Starting in 2010, the company committed internal resources to see how analytics could utilize “big data” to address a costly condition in the healthcare system – severe sepsis and septic shock (S4).

Sepsis is defined as an acute body wide inflammatory state (also called Systemic Inflammatory Response or SIRS) in the presence of a known or suspected infection. Sepsis causing organ dysfunction is called severe sepsis, and severe sepsis that leads to low blood pressure (hypotension) or insufficient blood flow is called septic shock (S4). Nationwide, sepsis and S4 affect hundreds of thousands of patients annually, particularly those in Intensive Care Units (ICUs), with a mortality rate of 30-60 percent.

In addition to the human cost, sepsis places a huge burden on hospitals. Annually, it costs hospitals an estimated $28 billion to $33 billion a year. A 2007 ruling by the Center for Medicare and Medicaid Services (CMS) limited payment to hospitals for certain preventable hospital-acquired infections.

In 2009, CMS added sepsis to the list of conditions covered by this ruling. As a result, the burden of cost of treating sepsis is gradually shifting to providers. Further, if private insurance companies, which typically model their guidelines after CMS, come out with similar policies in the future, the burden to providers will be compounded.

Among sepsis patients treated in the general ward setting, average total costs per patient are estimated at $13,900. Among sepsis patients treated in the ICU, average daily costs per patient are estimated at $29,900.[1]

As a result of findings like these, in 2008, physicians developed an approach for lowering the mortality rates due to severe sepsis and septic shock, which was called The Surviving Sepsis Campaign (SSC). Although clinical studies had been conducted in small and controlled settings, there was little evidence in large day-to-day clinical settings to support the efficacy of the Campaign’s interventions.

NEXT — The Sepsis Intervention Outcomes Project Detailed…>>>

 

Mercy Health, a St. Louis-based hospital system, shared our goal for designing better approaches to address severe sepsis and improve patient care. They had three objectives: harness the value of patient information to diagnose more quickly; decrease the time between official diagnosis and implementation of the standard of care; and lower mortality rates and overall health costs. Treating sepsis-related illnesses requires expensive resources and in the case of hospital acquired infections (HAI,) there may be challenges in reimbursement for that care

Sepsis Intervention Outcomes Research (SIOR) Project

To accomplish these goals, Mercy Health had many challenges and opportunities to consider. It had a large data set of EHRs containing patient data from several hospitals in the form of structured and unstructured data. System experts knew they needed to understand the efficacy of current treatment guidelines and its relationship to health outcomes, as well as what data they needed to diagnose sepsis earlier.

Working together, Booz Allen and Mercy Health created the Sepsis Intervention Outcomes Research (SIOR) project. This project sought to address two important research areas:

  • Compliance Analysis: To evaluate and measure effectiveness of hospital compliance with SSC guidelines for addressing severe sepsis and S4 by analyzing patient EHR records
  • Early Detection Analysis: To mine the EHR records for potential clinical indicators that could lead to early detection of S4

Booz Allen’s team led a cross-company effort that tapped analytical, clinical, administrative and business expertise. SIOR analyzed medical workers’ compliance with international standards of care for severe sepsis and S4, and compared that compliance with patient outcomes. Working together, Booz Allen and Mercy Health developed a unique analytical framework, including the creation of a unique temporal or time-based ontology, to allow for more efficient computation and discovery of underlying relationships.

For example, SIOR found that only 17 percent of hospitals followed best practices to treat S4 within six hours of diagnosis and that increased compliance to international guidelines for treatment is correlated with a decrease in patient mortality.

Mercy Health has already begun to use this information as a critical foundation for the implementation of an educational campaign to increase compliance to international guidelines. A direct correlation was found between a given hospital’s level of compliance and the health outcomes of severe sepsis/S4 patients – a roughly 20 percentage points decrease in mortality for the hospitals which most fully comply with the SSC guidelines versus the least compliant.

A second project sought to improve early detection of S4. Booz Allen’s advanced analytics experts helped develop an event-centric ontology (ECO) which incorporated natural language-processing (NLP) of medical personnel notes. ECO provided a formalized vocabulary and framework for evaluating electronic health records that speeded real-time discovery and harnessing of structured and unstructured data.

These capabilities enabled finding red flags for S4 in disparate sources that in turn could alert medical staff of potential situations earlier, buying precious treatment time and decreasing sepsis-related suffering and mortality. Booz Allen found an initial set of clinical indicators that may classify which patients with sepsis are at risk to develop S4. Though preliminary, these results could help to identify high-risk patients and prioritize their care.

Both of these ongoing projects provide insights into how hospitals can lower sepsis-related costs, freeing money for other patient care and other priorities using data. SIOR has already been a use case for advanced analytics to help fight other HAIs, identify potential discharge-readmission problems, and assist in a range of other disciplines.

Booz Allen and Mercy Health have developed a work stream for ECO implementation and tied its compliance analysis for the 6-hour and 24-hour standard of care guidelines to patient outcomes. The team has also validated the identified clinical indicators for early detection of S4.

During BIO IT World, Booz Allen will be presenting this case study to illustrate the promise of advanced data analytics and bioinformatics. We believe the vast amounts of patient data available through EHRs have created unprecedented opportunities to assess hospital-level compliance with SSC guidelines and detect early indicators of sepsis.

Ultimately, “big data” in healthcare will not be a storage or computation issue; it will be an asset to save lives.

Related Stories

Wrangling Big Data to Fight Pediatric Cancer

Predictive Modeling Meets Patient Behavior

Fixing Healthcare with Big Data

 


[1] Slade E, Tamber PS, Vincent JL, The Surviving Sepsis Campaign: raising awareness to reduce mortality, Crit Care. 2003; 7(1): 1–2.

Datanami