Posted on September 14, 2016
Sepsis is one of the most challenging conditions for healthcare systems around the globe and is a major concern to providers. In the United States alone, sepsis annually claims more than 258,000 lives, with some estimates putting the toll as high as 750,000 deaths. It remains both the deadliest condition treated in hospital critical care units and the most expensive. By 2013, the annual financial burden had climbed to almost $24 billion, an increase of $3.4 billion in two short years.
However, despite the overwhelming human and financial tolls of sepsis, efforts to improve outcomes often yield suboptimal results. Sepsis mimics common illnesses, so clinicians often initially do not recognize it, causing delays in treatment. Treatment delays are often fatal as the risk of death escalates by nearly eight percent every hour. The problem is twofold:
Sepsis is clinically suspected when a patient with an infection presents with two or more criteria for systemic inflammatory response syndrome (SIRS). However, most patients with SIRS symptoms don’t have sepsis. Hundreds of various comorbid conditions and medications can mimic SIRS in different patients. As such, most attempts to effectively use advanced CDS alerts often generate false-positive alerts, because they fail to achieve alert specificity rates higher than 50 percent, resulting in alert fatigue and ultimately, clinician rejection of CDS. For electronic alerts to impact outcomes, clinicians must be able to trust that the information they receive is accurate and improves the care they provide to patients.
To address the problem of alert fatigue, our team spent four years studying the variables that lead to missed cases or false positives and wrote a series of several hundred rules to improve the accuracy of sepsis alerts. Using an electronic surveillance system, patient data was aggregated and normalized into a cloud platform. Once in the cloud, the system used an algorithm to run the data against our sepsis rules to account for comorbid acute conditions, chronic diseases, medications and other associated parameters specific to each patient. Lab and vital sign parameter values and clinical parameters were also adjusted for the patient population.
Prior to its integration of CDS, Huntsville had already implemented a change management program designed to improve education for sepsis best practices and update protocols. To better support its sepsis program, the hospital also deployed POC Advisor. If positive results of sepsis are found, POC Advisor sends alerts to clinicians to ensure rapid treatment. These alerts can be accessed from a variety of devices, including mobile devices.
After a 10-month study period, results revealed a 53 percent reduction in sepsis mortality and unparalleled alert accuracy of 95 percent sensitivity and 82 percent specificity. Furthermore, sepsis-related 30-day readmissions dropped by 30 percent. These outcomes subsequently translate to Huntsville saving approximately 200 lives annually. The full details of these results are found in our JAMIA study, “Evaluating the impact of a computerized surveillance algorithm and decision support system on sepsis mortality.”
Based on our findings, electronic surveillance systems show significant progress in reducing the mortality rate and the expense associated with sepsis. Additionally, those systems that are regularly updated to reflect new guidelines and best practices, while regularly refining sepsis rules as new medical evidence is discovered to minimize alert fatigue, are likely to experience sustained improvements.