As the leading cause of hospital deaths in the United States, sepsis is a significant concern for patients and providers alike. In a 2020 study, the World Health Organization found a 46% mortality rate for patients treated in the ICU for sepsis. Sepsis care is expensive, costing the US healthcare system an estimated $33 billion per year. Data shows a single admission for sepsis care can cost anywhere between $22,000 and $70,000, making it the most expensive infection for hospitals, and reimbursement rarely covers those costs.
Many hospitals have dedicated teams focusing on improving sepsis care, and the COVID-19 pandemic has only magnified challenges for hospital leaders and sepsis coordinators. This leaves many care teams unsure about where to focus their sepsis efforts, both in saving precious lives and improving the bottom line.
Hospitals are turning to decision support systems based on artificial intelligence (AI) for help with the early identification of sepsis. However, experience has shown that AI is not the magic bullet it is sometimes made out to be. While AI technology holds promise at the point of initial diagnosis, purely AI-based sepsis identification tools rarely help care teams with one of the most challenging parts of sepsis – the ongoing patient management. To improve sepsis outcomes, hospitals must implement a sepsis program that leverages advanced, clinically validated technology for identifying sepsis as early as possible and use tools that help bedside care teams manage ongoing care in their fast-paced environments. Ultimately, technology that delivers a combination of early identification, continuous monitoring, and point-of-case suggested best practices for clinical management into the hands of committed and skilled clinicians, will help hospitals improve the quality and outcomes of care for sepsis patients.
Evaluating the full patient
AI-driven solutions hold significant promise for identifying sepsis early, but the technology must perform at consistently high levels of sensitivity, specificity, and seamless adaptation to different care settings. AI-based systems that produce high volumes of false alerts fail to engage clinicians, increase alert fatigue and worsen overall physician burnout; they do not improve patient care.
Sepsis identification technologies can improve sensitivity and specificity by mining and distilling relevant sepsis-related information from free-text clinical notes and unstructured ancillary narratives such as radiology reports via AI-driven natural language processing (NLP). Extracting pertinent clinical data from clinical notes provides the Sepsis algorithm with a comprehensive and holistic picture of the patient and superior sepsis identification and surveillance capabilities.
Supporting Precision Care
In addition to the valuable information often found in clinician notes, a successful sepsis management tool can, and should, account for the individual characteristics and demographics of the patient. When verified clinical considerations complement technology, more personalized care becomes possible, and hospitals can ensure patients receive the best care tailored precisely to their unique health profile.
Achieving Physician Engagement and Care Team Trust
Once a patient is diagnosed with sepsis, CMS guidelines define the compliant protocols for quality care. However, only about 50-60% of patients nationwide received care that was compliant with the CMS protocols (also known as CMS Sepsis Care Bundles). Bundle compliance is substantial as multiple studies have shown that when hospitals deliver care according to the bundle, sepsis survival rates increase, length of stay and readmissions are reduced, and the total cost of care decreases. The benefit of the sepsis bundle for care is clear, and bedside teams make every effort to deliver the best care according to the guidelines. However, hospitals can be chaotic and incredibly fast-paced. Even with care teams doing their best, without automated monitoring, patients can fall through the cracks. Technology can assist in the early identification of sepsis, but early identification alone will not improve outcomes.
A successful sepsis management program needs to monitor patient care over the duration of the patient’s stay and alert clinicians if a step in the patient’s care is being missed or if a patient’s health is further deteriorating. Systems focused exclusively on identifying patients with sepsis often lack the vital capability to monitor care over time. Technology that monitors patients over time and alerts when their care deviates from what is expected can make a significant difference in delivering the expected care and improve patient outcomes.
In addition, systems designed to optimize sepsis identification often fail to recognize when providers have already begun to treat a patient for sepsis and may send alerts that are just noise for the care team who are already delivering the proper care. Well-designed clinical decision support systems understand the dangers of over-alerting and are engineered to make discriminating decisions that minimize false alerts.
A holistic view of sepsis management
Delivering high-quality inpatient patient care has always been a team sport, with each team member playing a vital role. Sepsis care is no different, and many hospitals have added dedicated sepsis coordinators to monitor septic patients across the entire facility. Decision support tools need to adapt to this whole-house model and enable support team members to monitor patients across units and nudge a bedside care team when necessary to deliver optimal care.
In addition, sepsis committees and hospital administrators need an ongoing view of how the hospital performs in sepsis management. Regular access to sepsis-performance data allows them to understand if the existing care infrastructure and clinical interventions are making a difference in patient outcomes or not. Administrators can use these metrics to determine where they may need to focus their improvement efforts and where healthcare providers need extra support to better care for sepsis patients. With the ability to examine the process of sepsis care from the initial diagnosis to treatment outcomes, administrators can be appropriately equipped to narrow down where improvements should be prioritized.
Hospital leaders must be mindful of these considerations when selecting solutions. Ultimately, technology tools will be most helpful in improving sepsis outcomes when they offer early identification, decision support tools that are integrated into clinical workflows, and support for the entire care team. High-quality clinical decision support empowers teams to take the right action for each unique patient at the right time.
About Itay Klaz, MD, MHCI
Dr. Klaz is the Medical Director for Clinical Surveillance & Compliance at Wolters Kluwer, Health. He is responsible for directing clinical efforts toward developing, implementing, and supporting Wolters Kluwer’s suite of surveillance software solutions, primarily focusing on POC Advisor. Dr. Klaz is a clinical informatician, dermatologist, and former military surgeon.