The Big Data Revolution is Here
Can a Crisis in Critical Care be Far Behind?
By Christie Hentschel, staff writer
In the Age of big Data, Critical Data Needs Critical Care
In the coming decades, perhaps nothing will have a bigger impact on life than big data– and the growing number of new technologies that computationally analyze it to reveal patterns, trends, and associations that predict everything from what you may eat for dinner to where you bank and how likely you are to succumb to a cold. But even before a hospital is ready to leverage artificial intelligence (AI) or machine teaching (ML), its growing stores of data demand enhanced management, integration and communication to deliver their meaningful clinical, administrative and financial value. But is healthcare, and critical care, in particular really ready to leverage their full benefits?
Ushering in the Age of Data-Driven Care and Analytics
Of course, patient data has always played a major role in healthcare and continues to do so for every hospital, no matter where they are on the big data roadmap. Today many patients still wriggle uncomfortably in waiting rooms filling out lengthy paper forms, while others wish doctors would spend as much time looking them in the eyes as focusing on their computer screens.
Unquestionably for every facility, technology has increased the amount of data that can realistically be collected and stored. Some hospitals are already translating vast amounts of data into relevant clinical intelligence to enable fast, powerful potentially life-saving decisions, making them more effective thigh-tech healers.
Other facilities are also applying the analytics of early warning systems in critical care and other settings to look for cluster of symptom that can predict that an unstable patient is headed for a heart attack. More technologically advanced hospitals are developing data-based care guidelines that draw on the past experiences of vast pools of similar patients to help clinicians choose appropriate treatment paths for diseases from heart disease to AIDS.
Wherever a facility is on the data-driven care spectrum, its future will be in AI, ML and the technologies detail above etc. These technologies are poised to dramatically change health care.
How Good is Your Crystal Ball?
However, what many overtook is that advanced analytics are in effect high-tech, data-driven crystal balls—only as good as the information powering them. How good is that in the average hospital– and in critical care, with its multiple bedside devices monitoring the hospitals most vulnerable patients who may linger between life and death?
Unfortunately the medical system traditionally has been one of the last sectors to take advantage of the full benefits of modern digital technology, resulting a lesser data quantity or even flawed data that hampers the impact and perhaps even validity of the information emerging from these algorithms. In fact, AI may be more proficient predicting what a doctor will eat for dinner than what drug should be giving a patient.
A Siloed Data Crisis in Critical Care
Nowhere is this more evident – or is addressing it more important – than in the ICU and other high acuity units, where patients often cling to life, and the big data revolution is all but an empty phrase. Bedside biomedical devices typically provide the most vital and up-to-the minute information about patient condition as the primary, if not only source of continuous real time patient data. Potentially, this data that could be analyzed and used to drive real time. Yet biomedical devises have not evolved with the times to support the digital world.
Certainly, these devices excel at capturing body functions, creating as many as 300 data points every minute. But they typically operate in isolation, not connected to one another or to the larger healthcare IT infrastructure. Data remains siloed, inaccessible and is flushed away after 72 hours. If a clinician isn’t at the bedside to capture it, a piece of vital biomedical data plays no role in in patient care. In fact, today .01% of available patient data is used for clinical decision-making.
Historically , biomedical devices were built for clinicians who sat by the bed around the clock to monitor patients and periodically chart data. Given that no clinician sits beside 24/7 any longer, it is no exaggeration to say that critical data itself is in need of critical care.
Physicians and clinicians need more data, better data and open, accessible data. This enhanced clinical care data is needed not only in the department but hospital wide and beyond at scale as physicians and clinicians move through hectic days. More data, better data and open, accessible data is also is a prerequisite to leverage the true value of AI, ML and other advanced technologies today and in the future.
Achieving this means dramatically changing the way patients are monitored by unlocking, enriching and aggregating this data on a single, open digital platform. A system that collects, analyzes and openly communicates device and other critical care and patient data whether to a phone, at a virtual ICU or a remote location, will solve the problem. And it will solve the problem today while keeping hospitals poised technologies of the future.
The Critical Need for Enhanced Data in Critical Care
A closer examination of how critical care data is acquired and applied will help explain the pressing need to improve critical care data to enable dedicated physicians and courses working under some of the hospitals most challenging conditions to deliver extraordinary care.
The Need for More Data
While a vast amount of patient data is actually acquired in critical care, in today’s hospital environment, practically speaking physicians can access only limited portions for care decisions. Largely the issue is that they are viewing data through the EMR, which had made major contributions to clinical data access. However the EMR only store limited data types at discrete intervals with low sampling rates.
Enhanced care in high acuity units requires:
Access to continuous biomedical device data to leverage for care and use for real time analytics. Currently most hospital access this data through the EMR and it is store only Estimates are that biomedical devices collect more than 873,000 data points per hour yet only 6 of these are recorded in the patient record.
Access to continuous data for other data feeds vs. through the EMR. Sampling of other EMRs device data store only limited data types at discrete intervals with low sampling rates. Also Typically, EMRs are unable to capture information from some of the most critical devices such as IV pumps.
Data presented in clinical context. This information is typically lacking . Providing contexts is crucial to help clinicians understand the implications of information for a specific patient. In particular, the more relevant data caregivers have surrounding an event, the greater their understanding of the problem and more appropriately they can respond.
An interconnected system would enable greater context-awareness, which is often critical to determining whether a particular reading is — or is not — clinically significant: A heart rate of 170 on a treadmill test may warrant a low-priority condition whereas this same heart rate at an intensive-care monitoring station may be assigned a high priority.
The Need for Better Data
The quality of data presented to physicians also is a stumbling block to better care and will be a problem to the advanced data analyses that will drive care in the future.
Quality care in high acuity units requires:
Data classification. Clinicians are overwhelmed with data unrelated to the patient condition under consideration. A single ICU patient may generate 2,000 data points in a day. All this Information must be quickly and precisely classified and processed so that relevant information can be made available for timely and appropriate interventions. Add to that data from the numerous other information systems and data stores throughout a hospital and the importance—and challenge—of culling through all that information to eliminate irrelevant data comes into sharp focus. To act swiftly, decisively and with clinical confidence, presenting physicians with clinically relevant data is essential.
For example, in a modern ICU, a single patient can generate 2,000 data points per day, said Dr. Brian Pickering, an anesthesiologist and critical care physician at the Mayo Clinic in Rochester, Minn. In a 24-bed ICU like his, that’s 50,000 data points a day. Important information is easily lost, or forgotten.
Archiving of bedside device data for future analyses and trending. Bedside data is transient. Typically all physiological monitors and critical biomedical device data is deleted after 72 hours. This eliminates all the patient history, which is crucial for trending, examining the causes of a recent event or predicting the possibility of a future problem. Healthcare is flushing away its most valuable data
Integrated comprehensive visualization of data across devices on a single display. Bedside data is fragmented. Biomedical device information and relevant downstream IT systems lack any unified, synchronized visualization—forcing clinicians to interpret and reconcile a patchwork of dissimilar information on multiple device displays to form an accurate and meaningful patient picture, while racing against the clock to deliver needed care.
For example, a 2019 HIMSS survey found that 60% of responders felt that integration of clinical data systems will be their most significant challenge in delivering personalized care.
Because data is often manually input into the EMR, all data accessed through this system is subject to human error.
Critical care devices collect more than 873,000 data points per hour yet only 6 of these are recorded in the patient record. Typically, EMRs are unable to capture information from some of the most critical devices such as IV pumps. Moreover, information lacks context. Most significantly, data recording is also subject to a significant time lag. If information is manually keyed in, it is also subject to the possibility of human error Information input into the many EMRs and other hospital IT systems lacks standardization and allows free text. Variance in terminology makes data difficult to analyze, while unstructured language can only be analyzed only through complex natural language processing technology not widely used today.
Optimized alarms. Alarm parameters are not optimized and fail to provide context. While alarms provide crucial notification of patient problems, alarm fatigue from the unrelenting interruption of false and irrelevant alarms is the collateral damage. Optimizing parameters for patient populations and even individual beds to support more appropriate alarms and alerts will enhance clinician response. Providing information about the reason for the alarm helps clinician prioritize care and respond appropriately.
For example, an estimated 80–99% of ECG heart monitor alerts do not require clinical intervention. Additional evidence: As new devices are introduced, the number of alarms to which a healthcare professional may be exposed may be as high as 1,000 alarms per shift. The US Food and Drug Administration have reported over 500 alarm-related patient deaths in five years. The Joint Commission, recognizing the clinical significance of alarm fatigue, has made clinical alarm management a National Patient Safety Goal.
The Need for Open, Accessible Data
Anywhere, any time access to critical care device data in real time. Whether on phone or computer, visualization of critical device information is largely restricted to the bedside. While other patient data sets may be pushed to larger hospital IT infrastructure or mobile devices, access to biomedical device data, particularly crucial waveform data, is often limited to the p. It is typical done at the completion of a nursing shift. Point of care where it is viewed only intermittently by busy clinicians caring for multiple patients. It is also recorded at set intervals in the EMR but is no accomplished in real-time, Relying on hours old information, clinicians may delay treatment that could make an important difference in patient outcomes.
An estimated 420,000 ventilation patients have adverse events related to improper ventilator weaning. Effective weaning requires constant clinician monitoring and device adjustment, according to the Healthcare Cost and Utilization Project (HCUP) National Inpatient Database 2014.
Ability to share device data. Bedside device data typically is not directly shareable with other care team members. Device data is typically locked in proprietary formats and storage. While some information is sent to an EMR, accessing it from this system is not a desirable strategy, as detailed below.
Ancillary IT system data is siloed. While relevant data from other disparate downstream IT systems such as laboratory or pharmacy may be archived, typically it is in a proprietary format and not easily integrated or shared.
Hospital-wide, outmoded and circuitous communications paths compromise care. Central monitoring rooms for telemetry and other high risk patients, for example, often rely on legacy communications such as faxes and phone calls to convey problems to physicians and nurses at the beside. Modern IT infrastructure to make communications more timely, precise and transparent will support the availability of better patient care data.
Manual record-keeping like this is time intensive, prone to human error, and requires considerable effort for consistent long term execution.
Where Do We Go From Here?
To implement a successful precision medicine program in critical care, hospitals must solve the critical care data crisis. Only then can they meaningfully leverage today’s advanced data-driven healthcare tools. To accomplish this, hospitals need a single, centralized, vendor agnostic approach that can aggregate, neutralize and integrate all biomedical device data with other relevant patient data from the EMR and other systems across the enterprise and beyond. The hospital also needs features and tools to manipulate and analyze patient information with the added ability to communicate relevant information with clinical context in real time to providers wherever located. And they need a way to operationalize and deploy those analytics at scale across the hospital and healthcare system.