How Mayo Clinic Is Combating Information Overload in Critical Care Units

Executive Summary

Health care teams depend on electronic health records (EHRs) to compile important medical data from innumerable lab tests and medical devices, observations, treatments, and diagnostic codes. But in fast-paced critical care units, where even small errors can have big consequences, EHRs can overload physicians with information. The sheer volume of data in EHRs creates a staggering challenge in complex environments such as intensive care units (ICUs) and emergency medicine departments. Individual clinicians may have to sift through more than 50,000 data points to find key information. This proliferation of data (both meaningful and meaningless) and the workload created by EHR systems have been key drivers of clinician burnout and, paradoxically, introduced new threats to patient safety. What is more, relying only on EHR data greatly limits the insights derived from artificial intelligence algorithms or big data analytics. Mayo Clinic, the nation’s second-largest critical-care provider in the United States, with nearly 350 beds in 15 intensive care units (ICUs) across its campuses in Minnesota, Arizona, and Florida, decided to combat the data deluge with ambient intelligence: a set of decision-making tools powered by data on and insights into clinicians’ goals, work environments, strengths, and performance constraints. When layered on top of existing information infrastructure, ambient-intelligence applications can cut through the clutter and deliver the right information in a digestible form that clinicians can use, quickly and effectively at the patient’s bedside.

JUAN Diaz-Faes for HBR

But in fast-paced critical care units, where even small errors can have big consequences, this digital team member can overload physicians with information. The sheer volume of data in EHRs creates a staggering challenge in complex environments such as intensive care units (ICUs) and emergency medicine departments. Individual clinicians may have to sift through more than 50,000 data points to find key information. This proliferation of data (both meaningful and meaningless) and the workload created by EHR systems have been key drivers of clinician burnout and, paradoxically, introduced new threats to patient safety. What is more, relying only on EHR data greatly limits the insights derived from artificial intelligence algorithms or big data analytics.

Mayo Clinic, the nation’s second-largest critical-care provider in the United States, with nearly 350 beds in 15 intensive care units (ICUs) across its campuses in Minnesota, Arizona, and Florida, decided to combat the data deluge with ambient intelligence: a set of decision-making tools powered by data on and insights into clinicians’ goals, work environments, strengths, and performance constraints. When layered on top of existing information infrastructure, ambient-intelligence applications can cut through the clutter and deliver the right information in a digestible form that clinicians can use, quickly and effectively at the patient’s bedside.

Insight Center

We created a multidisciplinary team of clinicians, researchers, and experts in clinical informatics to design and test information-technology tools that can help, rather than hinder, clinical care. The ambient-intelligence approach we adopted prioritized a deep understanding of clinicians, the way they work, and the environmental factors they face. Using a NASA Task Load Index, we identified clinicians with a very high mental, or cognitive, workload who continuously have to filter important information out of the cluttered environment.

Subsequently, over a two-year period, we conducted 1,500 interviews with clinicians from Mayo Clinic ICUs nationwide. With these insights, we identified that out of tens of thousands of pieces of data pouring through EHR, roughly only 60 pieces are crucial patient information that clinicians needed to access quickly and easily for effective care. This information included both expected data points, such as blood pressure and medications, as well as less obvious but critical information such as cough strength or previous difficulty with endotracheal intubation.

Next, we needed to find a better way to deliver the crucial information to clinicians at the point of care. We built an EHR interface for clinicians in the ICU called Ambient Warning and Response Evaluation (AWARE), which we introduced in our ICUs in Rochester, Minnesota, in 2012, and in our campuses in Phoenix/Scottsdale and Jacksonville, Florida, in 2014. A rules-based, ambient-intelligence application, AWARE filters out the meaningless data and delivers context-specific, high-value information to clinicians in real time. It contains over 1,000 rules that run continuously through data and enrich it with insights from clinicians and patients. Data is organized around familiar clinical concepts needed for timely and accurate decision-making.

For example, the conventional EMR displays are cluttered with irrelevant data, making it easy to miss changes in hemoglobin, platelets, and coagulation factors — all critical for recognizing and treating acute bleeding. By prioritizing these data elements on the dashboard with color coding that indicates severity and urgency of specific corrective intervention, AWARE displays allow instantaneous recognition of an ICU patient at risk of severe bleeding complications.

AWARE provides a real-time overview of every ICU in the Mayo Clinic system, using visual displays that make it simple to scan and identify patients in need of urgent interventions. Each patient is represented by a square with icons that represent the status of required tests, scans, and procedures. It provides a quick picture of overall acuity, and clinicians can drill down into each patient’s data — down to each organ system.

While conventional alarms and EHR alerts create often create meaningless noise that gets lost in the hustle and bustle, AWARE smart alerts are integrated into the clinician’s workflow. They notify the clinician of potential omissions only if the clinician’s actions do not match the patient condition, minimizing the chance of harmful interruptions.

For example, our “VILI (ventilator-induced lung injury) Sniffer” provides an automated surveillance of mechanically ventilated patients that notifies providers only if the ventilator settings do not match the patient condition based on gender, height, and the presence of acute respiratory distress syndrome. Another application, “Sepsis Dart,” continuously surveys patients in emergency departments and medical intensive care units for the timely and accurate implementation of best practices (cultures, lactate, antibiotics, fluid) for diagnosing and treating sepsis. Yet another example is context-specific smart checklists such as CERTAIN (Checklist for Early Recognition of Acute Illness) that focus on completing common processes of care that are sometimes overlooked or missed in busy environments.

Compared to standard EHR interfaces, AWARE improves the cognitive performance, efficiency, and reliability of human decision makers. It also saves three to five minutes on chart review per patient per day. With an average ICU clinician workload of 15 patients per day, the savings mean that more than one hour of additional clinician time can be devoted to patient bedsides, improving often-inadequate shared decision-making. In a subsequent study, AWARE implementation was associated with improved patient outcomes and reduced costs in the ICU. Adjusted for illness severity, the odds for hospital mortality of critically ill patients treated after AWARE implementation were reduced by half (odds ratio 0.45, 95% confidence interval 0.30 to 0.70). In addition, the length of ICU stay decreased by 50%, length of hospital stay by 37%, and total charges for hospital stay by 30% ($43,745 per hospital admission).

As an example of ambient-intelligence applications used in the emergency and ICU settings, AWARE delivers results that clearly suggest the importance of human insights and creativity in developing information technology for clinicians. Advancing ambient-intelligence applications and combined human–computer partnerships has the potential to help solve some of our most meaningful and challenging health care problems.

Disclosure: Mayo Clinic Ventures licensed part of the technology referenced in this article to Ambient Clinical Analytics, which sells clinical decision-support and alerting tools to hospitals and critical care providers. Mayo Clinic and the authors have financial interests in Ambient Analytics. Mayo Clinic uses any revenue it receives to support its not-for-profit mission in patient care, education, and research.