Game changing approach advances well beyond predictive analytics.
The state of sepsis is staggering: It is a leading cause of death worldwide. More than one million Americans develop severe sepsis each year. Sepsis is the most expensive condition treated in the United States’ healthcare system.
Care providers have very poor insight into sepsis
Even though there is a spotlight on this serious condition, the reality is that hospitals struggle to recognize and combat sepsis in a timely manner. For decades treatment of sepsis was reactionary. Clinicians waited for telltale signs of progressive infection such as tachycardia, tachypnea, or hypotension before initiating treatment. Over time it became clear that sepsis was not a fait accompli as proactive steps taken by the clinical team could mitigate or prevent significant morbidity. The current trend in sepsis care has been to predict who is at high risk for sepsis before the telltale signs develop. Risk scores such as SOFA (Sepsis-related Organ Failure Assessment) have been developed in an effort to accurately predict sepsis risk. However, these results have limited utility due to their insensitivity.
A major challenge in sepsis avoidance and detection is that care providers are always a step behind. This is not their fault. Front-line care teams are expected to leverage a combination of diagnostic and treatment protocols to diagnose and enact early sepsis treatment. This typically requires synthesizing hundreds of data points on a daily basis without the ability to identify which are actually causally related to the diagnosis of sepsis. These analyses do not reflect minute-to-minute changes in patients’ conditions, are not sophisticated enough to consider the many possible causal variables, and they burden time-constrained providers with interpretation.
Additionally, sepsis detection and treatment is only as strong as the experience of a care provider and his or her access to and understanding of an ever-evolving universe of clinical trial data and expert guidelines.
“We have decades of learning about sepsis. Despite all of our knowledge and current technology, a quarter of a million people still die every year because we are not employing tools to truly identify, in real-time, who is at risk,” said Adam Silverman, M.D., a member of the Life2 Clinical Advisory Team. “Our current assessment and treatment models have not moved the needle. We need to link the power of enterprise-grade artificial intelligence and machine learning with sepsis care delivery. These technologies expand the training set that a clinician is exposed to and provide knowledge and experience that would take decades for an individual clinician to obtain. This enables every provider to make expert care decisions. In addition, the predictive capabilities of these technologies are far superior to any tool that currently exists.”
A new approach to patient-level sepsis prediction, prevention and care
We are all tired of market buzzwords, trends and solutions that burden care teams and do not really make an impact. It is time for a new approach. Artificial intelligence (AI) represents a paradigm shift in the care of sepsis by allowing the development of a highly-personalized digital fingerprint for sepsis for every patient. When an individual patient’s sepsis fingerprint is compared to thousands of similarly diagnosed and treated patients, we can determine their probability of becoming septic. When clinicians are armed with earlier clues about an individual patient’s risk of sepsis, they can improve treatment time, and design more effective and personalized care plans.
What does this digital fingerprint show care teams?
- Real-time, patient-level sepsis risk, typically hours but even days before current identification methods. When hospitals have the analytics horsepower to continuously search for causal relationships in evolving clinical data, the underpinnings of sepsis are visible much earlier. This analysis can alert care teams to real-time sepsis risk, accelerate the process of differential diagnosis development, and predict a patient’s decline earlier than the clinical team would otherwise detect it. This will impact sepsis outcomes in ways the have been previously impossible.
- How to personalize expert guidance to enact more effective sepsis treatment plans. No two sepsis patients are the same. Machine learning algorithms that comb every hospital system can create dynamic recommendations that are not only rooted in the latest clinical guidance but also influenced by real-time patient status. When these suggested actions are delivered to care teams at the bedside, along with data on the historic effects of choosing or rejecting them, they are empowered to start accurate and effective interventions that have the best chance of improving the patient’s outcome.
- Empirical insights on what is working and not working to combat sepsis. AI and machine learning have advanced beyond simply making predictions. Capabilities now include continuously evaluating outcomes based on clinicians’ responses to suggested actions. Technology can provide clinical teams with historic performance for acceptance or rejection of specific, evidence-based treatment recommendations. This lets the clinical team see exactly what has and has not worked in the treatment of sepsis patients. Hospitals can mine their data for more meaningful clues and insights on how it is managing sepsis. These insights can be incorporated into ongoing physician practice evaluation and education to continue to improve sepsis care delivery.
One hospital’s results
At a large, acute care hospital in Georgia, early results of an AI-driven approach to sepsis care have been remarkable. The Life2 AI Platform is currently predicting 62% of sepsis cases before clinical teams recognize them. The mean time of prediction before clinical diagnosis is six hours, a valuable window of time to carry out earlier intervention for sepsis care. Its previous sepsis identification method had a sensitivity of 61%, whereas Life2 identifies 91% of sepsis cases.
Providers at this hospital now also have a clinical playbook for every sepsis patient. This includes the established guidelines for treatment combined with unique, patient-level data and insights that ensure treatment is personalized for every patient.
Lastly, insights produced by the Life2 AI Platform make it possible for the hospital to improve clinical education. It stores detailed insights on intervention adherence and the impact on mortality and length of stay. It also provides new insights, including data on clinicians’ continued use of outdated treatments.
The right analytics and data strategy is critical
Hospitals already hold the raw data to preventing sepsis. They need more powerful technologies to decipher it.
“We need more than just clipboards and sticky notes, which have only made small, incremental impacts in the last two decades. The Life2 platform helps clinicians impact sepsis outcomes in ways that were not previously viable,” said Sam Wilkes, Founder, Chairman, and Chief Executive of Life2.
Life2 is an AI-powered risk mitigation platform built for complex healthcare enterprises. It has mapped the known universe of clinical and operational risks facing acute care hospitals, and combines that vast intelligence with the ability to analyze data feeds from any information system to identify real-time causal patterns and alert care teams to patient-risk. Life2’s cutting-edge AI can continuously process billions of data elements to provide valuable insights for the endless combinations of DRGs, comorbidities and operational risks facing hospitals. The end results are optimized patient outcomes, quality metrics and financial performance.
In the case of sepsis, the Life2 platform knows every patient’s unique, real-time risk from the time they are admitted and at every minute throughout their course of care. It alerts care providers to clinical decline and provides the corrective action that should be applied to that particular patient at that particular time to course correct. This same approach can be applied to mitigate risk in areas such as pneumonia, diabetes, congestive heart failure, length of stay, readmissions, reimbursement, quality metrics, safety goals, and resource utilization.
Everyone is talking about applying AI and machine learning to solve the healthcare industry’s problems, but not all solutions are equal and, unfortunately, most do not have the deep technology architecture or clinical workflow to deliver on their claims. Care teams do not need another analytics system that tells them what happened or what is happening; they need a tool that gives them an edge in improving patient care by telling them what will happen with a specific patient, and what actions have been proven to mitigate that unique risk.