Assessing Critical Illness During Emergency Medical Services Care

Emergency medical services (EMS) and systems in the United States expanded considerably after legislation in the 1960s seeking to address the variable and often substandard out-of-hospital care of the time.^{1 - 3} Although the initial focus was on out-of-hospital treatment of major trauma and cardiopulmonary arrest, the movement laid the foundation for the expanded role of EMS in the current health care system. The broadened role of EMS in the contemporary care of patients who experience trauma, cardiac arrest, myocardial infarction, stroke, and shock highlights that EMS personnel are often the first to recognize and care for the sickest patients.^{4 - 9}

Despite substantial progress in development of EMS and significantly increased capabilities of EMS practitioners, the optimal intersection between EMS care and hospital care remains uncertain. The 2007 Institute of Medicine report that assessed the status and future of emergency health care in the United States dedicated 1 subcommittee to the vital role of EMS care.¹⁰ Among many proposals, the report called for enhanced coordination between out-of-hospital care and hospital care, centering regional care for select conditions around specialty or high-volume hospitals. Acute trauma and ST-segment elevation myocardial infarction are examples for which established triage guidelines, regional organization, and tight connections between EMS and dedicated hospital services synergistically improve outcomes.^{4 ,8} However, these conditions represent a fraction of the critical illnesses treated by EMS personnel. Better recognition and organization of care also is necessary for the remaining sizable group of patients with other severe illnesses, such as severe sepsis, respiratory failure, acute decompensated heart failure, and hemorrhagic shock.

In this issue of JAMA, Seymour and colleagues¹¹ report on the creation of a clinical decision rule based on EMS clinical data to predict development of critical illness. Using a detailed cohort assembled from the mature King County, Washington, EMS system and limiting the analysis to adults without traumatic conditions or cardiac arrest, the authors developed a rule designed to identify the presence or impending development of critical illness, defined as patients who died, needed intubation, or met severe sepsis criteria during the hospital stay. In the context of regionalized care, such a tool would fill a void, assisting EMS practitioners with identifying the sickest patients. The resulting improved triage could help EMS personnel choose the right destination care at the right time for the right patient.

The best clinical decision rules are created with sound methods, have face validity, improve care, and are practical to apply.¹² In the out-of-hospital environment, distinct characteristics affect the creation and use of any clinical decision rule. EMS personnel vary widely in training and background, and may include basic emergency medical technicians (EMTs), advanced paramedics, and, in some prehospital systems, nurses and physicians. EMS systems have different sizes and configurations, including ground vs air medical agencies and volunteer vs paid or municipal groups. EMTs and paramedics receive less training than physicians and nurses yet must care for severely ill patients in uncontrolled and challenging settings. In addition, the EMS care interval is much shorter compared with other settings, often involving rapid assessment and decision making within minutes. Any successful out-of-hospital clinical decision rule must be amenable to these conditions.

Does the decision rule presented by Seymour et al¹¹ meet these needs of an EMS clinical decision rule? The authors used a clear set of predictors of critical illness (ie, older age, lower systolic blood pressure, abnormal respiratory rate, lower Glasgow Coma Scale score, lower pulse oximetry, and nursing home residence) and a composite outcome (death, intubation, and severe sepsis during hospitalization), all seemingly simple and measurable. However, even these simple predictors are often not obtained or are inaccurately assessed (eg, ambient vs supplemental oxygen saturation measurements, clearly defining nursing home), threatening the ultimate utility of the rule. Similarly, the nonfatal outcomes may not reflect all of the markers of critical illness. For example, patients with massive bleeding or cardiogenic shock may not experience these events, yet they often require critical care to avoid poor outcomes. Moreover, analyses of outcomes in critically ill patients should examine death separately from nonfatal events and at both short and longer intervals (eg, in hospital and at 60 or 90 days). This approach allows enhanced detection of deaths potentially related to the illness, recognizing that many factors influence longer-term outcomes.

The model reported by Seymour et al consisting of 6 clinical out-of-hospital data elements that create a score between 0 and 8 calibrated well with the outcomes, with critical illness occurring more commonly among patients with higher score strata. As expected, older age, hypotension, abnormal respiratory rate, abnormal sensorium (using the Glasgow Coma Scale score), hypoxemia, and nursing home residence (a common surrogate for other comorbid conditions) identified patients who developed critical illness during hospitalization. However, it is uncertain whether the rule actually predicts subsequent critical illness or merely categorizes the manifestations of critical illness. Moreover, there is no evidence to help understand how this critical illness decision rule would add to current EMS practice. For example, well-trained paramedics may already use these factors as part of current clinical gestalt. Finding a process by which any out-of-hospital EMS provider could rapidly and reliably assess and act on a 0- to 8-point scoring system is a challenge needing a solution for this or any similar rule.

The prediction rule reported by Seymour et al requires refinement and further validation before it can be used to aid regionalization efforts. At the recommended cut point of 4 or more points, the rule has a sensitivity of 22%, a performance that is less than what most EMS personnel, system leaders, and health care experts would deem acceptable. Compounding this problem is that 64% of patients scoring 4 or more points ultimately did not require critical care, potentially diluting opportunities to concentrate critical care at designated centers. Altering the score threshold to improve sensitivity worsens the mistriage rate by adding to the frequency of false-positive classifications.

Even if rule performance could be enhanced, to which regional center should EMS personnel transport high-risk or critically ill patients? The ability to provide optimal care for patients with sepsis, respiratory failure, hemorrhagic shock, or other critical illness can vary within and between hospitals, yet the broad-based rule reported by Seymour et al does not aid in choosing among sites. Some centers have multiple specialty capabilities, but an unresolved issue is whether a comprehensive critical care center would necessarily deliver better care and achieve better outcomes than a center focused on fewer or a single condition. An essential requisite to deploying any out-of-hospital critical illness triage tool is an evidence base delineating which conditions to regionalize, what treatments to provide, and how to organize this care. Despite broad recommendations, the current data formally supporting regionalization of critical and specialty care remain limited.^{10 ,13 - 15} Even if all of these questions are resolved, autonomy concerns also would need to be addressed: would patients want regionalized care or would they prefer the best care nearest to their home or where they have received care previously?

Despite these limitations, the findings of the study by Seymour et al have potential value for EMS systems and care. First, the prediction rule using out-of-hospital clinical factors can help assess illness burden when analyzing existing data sets or seeking balance between groups before an intervention. Second, this effort offers an initial framework for risk identification in a challenging environment. Perhaps other data, such as body temperature, lactate or other physiologic measurements, specific physical examination, and historical features, could improve the performance of this or any rule. With improved information technology, the potential exists to extract key data and deliver it promptly to out-of-hospital providers and emergency department or critical care personnel at receiving hospitals. Most importantly, this work should prompt reevaluation of the broader goals and organization of community critical care systems.

Seymour et al have highlighted the interdisciplinary spectrum of critical care, the pivotal role of EMS in this continuum, and the opportunities for classification and care at the earliest out-of-hospital stages. The challenge is to find better ways to deliver care to those most in need, starting as early as possible.