Early identification and treatment are key to reducing the high cost of sepsis care and improving patient outcomes. Accurate sepsis alerts are key to driving improvement.
Frequently, a hospital would manage sepsis care using the EHR's generic sepsis tools and alerts, aiming to accelerate accurate diagnosis and treatment. Unfortunately, the sepsis alert results are marginal at best.
Sepsis is complex and requires dedication to education and change management
Sepsis is a clinical diagnosis. Delivering evidence-based care to address the rapid change in biomarkers is the goal. Changing outcomes requires a hospital's commitment to education and change management, adopting evidence-based consensus guidelines and tools, and compliance with measures from the Center for Disease Control (CDC) and other governing bodies. Not only do you have to prevent diagnostic errors, but clinicians also must make the diagnosis quickly and early to save lives.
Where EHRs fall short in identifying and alerting for sepsis with SIRS alerts
EHR sepsis alerting systems commonly utilize Systemic Inflammatory Response Syndrome, or SIRS alert criteria, for sepsis surveillance. The sepsis detection system will trigger a "sepsis alert" if the EHR identifies SIRS criteria.
Commonly seen false positive alerts – SIRS alerts
Unfortunately, SIRS-based alerts that do not account for the patient's co-morbidities, medical and surgical history, medications, and demographics are often false positive, unreliable, and disruptive. Positive SIRS criteria are present in many hospitalized patients who do not develop sepsis, and frequent false alerts tend to be ignored by clinicians. For truly septic patients, the resulting delay in antimicrobial treatment and hemodynamic support at the onset of severe sepsis increases mortality rates by 7–10% per delayed hour.
Alternately, sometimes EHR systems deploy alerting systems using the Modified Early Warning System (MEWS) instead of SIRS. Regardless, both SIRS and MEWS-based alerting systems are known to be sensitive but have poor specificity.
Today, some systems are starting to use the Sequential Organ Failure Assessment (SOFA) and qSOFA scoring systems for alerting, but these newer systems are already known to have poor sensitivity and a delay in identifying abnormalities.
Because of EHRs’ lack of sensitivity and specificity in alerting, each fulfillment of additional sepsis criteria results in the repeat activation of a sepsis alert leading to:
Lack of accurate sepsis alerts
Clinicians' complaints about alert accuracy are warranted since they may be linked to significant patient hazards. A well-cited study revealed that clinicians ignore EHR safety notifications between 49 percent and 96 percent of the time because of alert fatigue. Computerized Physician Order Entry (CPOE) systems in the EHR generate warnings for 3%–6% of all orders entered, meaning that a physician could easily receive dozens of warnings each day—putting patients at risk if accurate alerts are ignored as well.
Clinical alert fatigue due to lack of clinical relevancy
Early identification of sepsis alerts must be specific and informative. Alert fatigue occurs when false alerts for non-septic patients are presented to the clinicians. They discredit the credibility and reliability of the sepsis decision support system. SIRS-only alerts that do not account for co-morbidities, medications, patient's history, and lack of suspected infection source have low specificity since many patients with unrelated conditions may present with abnormal vital signs and trigger the alert. Positive SIRS criteria include abnormal temperature, respiratory rate, heart rate, and white blood cell counts and can be linked with thousands of non-sepsis medical conditions. With typical EHR alerting, clinicians ignore ~95% of notifications as they're typically wrong more than 90% of the time due to the 15% specificity of SIRS-based platforms. Only 1 in 10 patients with alerts will have sepsis! Improving clinician utilization of any system is dependent on being able to trust the alert content delivered.
Alert fatigue can be drastically reduced by implementing clinical algorithms to rule our non-sepsis conditions with positive SIRS criteria and early identification of suspected infection sources. Clinicians often document the relevant data in textual narratives, such as progress notes, assessments, radiology reports, and consulting notes. While speaking on a panel titled “AI & Sepsis: How Can It Inform Bedside Care?”, Cerner Vice President Tanuj K. Gupta, MD, MBA, referred to estimates that some 60 percent of potentially clinically relevant information in healthcare appears in unstructured data, while others assess that rate to be up to 80%.
Using Natural Language Processing (NLP) with AI capabilities, surveillance solutions can extract real-time information from clinical documents and provide crucial and comprehensive clinical data. Utilizing these advanced technologies, like NLP, significantly increases the accuracy and specificity of the generated alerts, reducing the system's noise-to-signal ratio. It engages the clinicians with the alerts and evidenced-based clinical recommendations that are provided.
Lack of timely alerts negatively impacts patient outcomes and mortality
Often EHR alerts are fired late or miss the early signs of sepsis, thereby missing critical opportunities to deliver early and rapidly effective treatment for septic patients. If the sepsis detection criteria used by the system require evidence of organ dysfunction (for example, by following the SEP-3 protocol) many early-sepsis cases will not trigger alerts early. Alerts would trigger only once the patient has already deteriorated to the advanced, life-threatening stages of severe septic shock, severe organ dysfunction, permanent organ failure, and eventually death.
Surveillance technology + process + people can move the needle on sepsis care
Although EHR vendors have developed sepsis alerting that appears to be a bargain since it is a "free" product, this has not proven to be the case. If one takes a closer look, in-house healthcare teams spend a great deal of energy and time on implementing and integrating the surveillance solution into their workflows in an (unsuccessful) attempt to improve its accuracy, costing time, and significant clinical resources. There is a better answer. Third-party surveillance technology used with an EHR can offer benefits to offset these gaps.
Surveillance technology, like Sentri7® Sepsis Monitor, which focuses on accurate and timely alerting, has moved the dial on detection and treatment by drawing on built-in intelligent clinical content and rules and incorporating change management services for hospitals. The core decision engine includes thousands of clinical algorithms developed by subject matter experts and scientific best practices. When accurate enough to account for a wide range of sophisticated clinical scenarios and relevant comorbid medical conditions, surveillance systems that deliver decision support to the point of care can improve outcomes where EHR-based systems have failed. Plus, centralized monitoring helps nursing leadership and sepsis coordinators manage patients across the hospital to the SEP-1 requirements.
A 2016 study evaluated the impact of a computerized surveillance algorithm and decision support system such as this on sepsis mortality. The study assessed - Sepsis Monitor, formerly POC Advisor, - a highly accurate system that used a combination of change management, computerized surveillance, and mobile point of care alerting to:
- decrease sepsis mortality by 53% in hospital units where the sepsis initiative was implemented
- improve alert sensitivity to 95% (effective screening tool) and the specificity to 82% (the sepsis cases were real) when compared to the gold standard of physician chart review
- firing early alerts in the course of sepsis so that rapid and effective treatment algorithms could be implemented
- improve clinician utilization by providing a Clinical Decision Support system that incorporates the complexities of a patient’s presentation
A highly-complex condition to diagnose and treat, sepsis is a prime target for technological intervention beyond the EHR. However, as is true in most forms of health IT, not all solutions are equal. Despite the early nature of this market, there are solutions, like Sepsis Monitor, that can support care teams in delivering improved sepsis outcomes, including mortality, morbidity, and cost of care.
