Risk-score Based ICU Triage

Multiple recent studies have demonstrated considerable between-hospital variability in ICU utilization for stable patients with NSTEMI and a lack of association between higher hospital-level ICU utilization and short-term mortality. Moreover, severity of illness, as measured by a traditional in-hospital mortality risk score, has only a trivial correlation with ICU utilization. A minority of initially stable patients with NSTEMI (\~15%) deteriorates clinically while hospitalized and requires ICU care for management of cardiac arrest, shock, arrhythmias requiring pacing, stroke, or respiratory failure. Across a variety of conditions outcomes are better when patients are admitted directly to the ICU from the emergency department (ED) rather than transferred in after admission. However, the cost of caring for patients in the ICU is substantially more than the cost of caring for these patients in a non-ICU environment. Furthermore, treating patients that do not require intensive care in the ICU exposes them to unnecessary risks of ICU care, including medication errors, adverse procedural outcomes, delirium, and excessive noise. Reducing ICU utilization for stable patients with NSTEMI may reduce costs and improve patient satisfaction.

Using data from a nationally-representative registry enrolling patients with acute MI, the investigators developed the ACTION ICU risk score. Incorporating demographic, clinical, and laboratory data obtained routinely in the ED work-up of patients with suspected acute MI, the ACTION ICU risk score calculates the risk of in-hospital complications mandating ICU care for initially stable patients with NSTEMI. Complications mandating ICU care were defined as death, shock (cardiogenic or otherwise), cardiac arrest, high degree heart block requiring pacemaker placement, respiratory failure, or stroke. The risk score's c-statistic was 0.72, indicating good discrimination. Importantly, it identified > 50% of patients as being at \< 10% risk of in-hospital complications mandating ICU care.

However, the clinical and financial implications of using this score to guide ICU triage in routine clinical practice are unknown, and the risk score has not been prospectively validated.

The investigators will create a calculator for the electronic health record that automatically calculates the ACTION ICU risk score for all patients with NSTEMI, as identified by their initial troponin value. Once the score is calculated, it will provide the score, and the patient's risk of clinical deterioration to the ED physician, along with a recommendation for where patients at that risk should be treated. The ED physician, working with the cardiologist on call, will then decide where the patient should be treated.

After one year, each patient for whom the score was calculated will be identified by a query of the electronic medical record. From the electronic medical record, the investigators will identify whether the patient was initially admitted to the ICU or to a non-ICU unit, whether the patient was transferred to the ICU during their hospital course, and whether the patient had clinical complications mandating ICU care (death, shock, cardiac arrest, heart block requiring pacemaker, stroke, or respiratory failure). The investigators will also compare total hospital costs for caring for NSTEMI patients before and after roll-out of the ACTION ICU score electronic medical record plug-in. Study completion will be defined by the last date of data extracted from the medical records for these patients.