We conducted a retrospective analysis of eight years of data from the Clinical Outcomes in Pulmonary Embolism Research Registry (COPERR), an observational registry of adult patients treated by the CODE PE team across 20 emergency departments (EDs) within the Atrium Health system in North Carolina, USA. The registry and related reports were approved by the local institutional review board. We examined patient characteristics, process outcomes, and clinical outcomes. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines [10]. The only use of large language models (i.e., ChatGPT, Microsoft 365 Copilot) was to paraphrase our original content to provide more succinct explanations of a few concepts or to organize tables.

Three Atrium Health hospitals (Hospitals A, B, and C) serve as PE referral centers based on their 24/7 availability of catheter-directed intervention (CDI), dedicated CODE PE teams, in-house critical care services equipped to manage high-acuity cases, and ability to accept transfers at any time. All three can initiate extracorporeal membrane oxygenation (ECMO) cannulation, but ongoing ECMO management requires transfer to Hospital B, the academic center. Non-referral centers are Atrium Health EDs not located within Hospital A, B or C. One non-referral center (Atrium Health University City) had partial CDI support from vascular surgery and interventional radiology between 2017 and March 2023. After that, patients with massive or severe submassive PE were redirected to the referral centers for advanced care.

We included all adults (> 18 years) with confirmed PE for whom CODE PE was activated in the ED between January 2017 and August 2024. All had evidence of right ventricular (RV) strain, defined by an RV to left ventricle ratio (RV: LV) ≥ 1.0 on computed tomography pulmonary angiography (CTPA) or signs of RVD on echocardiography (point-of-care or formal). We defined PE severity by a hybrid of classifications used by the PERT Consortium, American Heart Association, and the European Society of Cardiology [11,12,13,14]. Patients with RVD and sustained hypotension (systolic blood pressure < 90 mmHg for 15 or more minutes), vasopressor use, or cardiac arrest were classified as massive PE. Those with RVD but without criteria for massive PE classification were classified as submassive PE, and we segmented this group into severe submassive PE and non-severe submassive PE as follows: 1) Those with RVD and episodic hypotension (systolic blood pressure < 90 mmHg for less than 15 min), shock index > 1.0, respiratory distress, syncope/presyncope, or clot-in-transit were classified as severe submassive PE. Patients without massive or severe submassive features were considered non-severe submassive PE. Our CODE PE program has distinguished subgroups of submassive PE since March 2020. For CODE PE activations with submassive PE prior to March 2020, we retrospectively assigned the appropriate subgroup (severe submassive vs. non-severe submassive PE). In March 2023, an internal retrospective review of COPERR data by a multidisciplinary committee found patients who met the severe submassive PE definition above were more likely to deteriorate from submassive to massive PE or need rescue intervention. This classification has since informed real-time triage within our system [15, 16]. We excluded patients diagnosed with PE during hospitalization as these cases are not entered into COPERR.

Our multidisciplinary CODE PE team includes the ED physician and an interventional specialist—either from interventional cardiology, interventional radiology, or vascular surgery—depending on available resources at each site. In more complex cases, cardiothoracic surgeons and ECMO specialists may also be involved. Vascular medicine physicians contribute to post-acute care, including follow-up planning and long-term anticoagulation management. At non-referral centers, the team is supported by a virtual critical care physician (VCC), a board-certified intensivist available 24/7 from a centralized location. The VCC verifies PE severity status and determines if the patient needs to be transferred to a referral center and, if so, with what level of urgency. Initial decisions on CODE PE activations are made jointly by the VCC and the initiating ED physician, with additional specialists consulted in real time as needed.

Atrium Health ED physicians and CODE PE team members receive CODE PE committee-led education, which is supported by CODE PE risk stratification algorithm posters, standardized order sets and power plans within the electronic medical record (EMR), rigorous data capture, and regular quality improvement reporting to the CODE PE committee.

As illustrated in Fig. 1, CODE PE activation requires confirmation of PE on CTPA along with evidence of RVD, defined as either elevated troponin or an RV size greater than the LV on imaging (CTPA, comprehensive transthoracic echocardiography [TTE], or point-of-care ultrasound). [Trained ED physicians performed point-of-care echocardiographic visual assessments for RVD [17, 18]. Sonographers performed comprehensive echocardiographic assessments and measurements.] Elevated troponin was defined using i-STAT cardiac troponin I or T > 0.07 ng/mL or high-sensitivity troponin levels exceeding 12 ng/L for females or 20 ng/L for males. TTE reporting included focused assessments of RV dimensions, systolic function, pressure, and cardiac output. Other CODE PE guidelines were internally decided. For example, when restarting heparin after systemic thrombolysis administration, we used a cut-off of 110 s to signal approach into the therapeutic range (60–100 s) and restart of heparin.

CODE PE pathway in use March 2023 to present

When an ED provider in a referral center activates CODE PE, a secure in-house chat system simultaneously notifies all relevant CODE PE team members. When an ED provider in a non-referral center activates CODE PE, in-house multidisciplinary expertise is not immediately available, so the secure chat only notifies the VCC. The VCC then coordinates rapid triage, verifies PE severity, consults with specialists, and facilitates transfers for patients needing advanced care.

As shown in Fig. 2, CODE PE activations between January 2017 and March 2023 included patients with massive, severe submassive, and non-severe submassive PE. In March 2023, the activation criteria were revised as part of a quality improvement initiative to exclude non-severe submassive PE since these patients rarely required escalation or intervention and were effectively managed by primary teams. This change, described previously, improved clarity and consistency across the health system.

CODE PE pathway in use January 2017 to March 2023

Trained data extractors reviewed each patient’s EMR for demographics, PE risk factors, comorbidities (including the Charlson Comorbidity Index [19]), clinical presentation, process outcomes, and clinical outcomes (in-hospital and 30-day post-diagnosis). For time-based measures, time-stamps were abstracted from the EMR for radiologist signature on computed tomography interpretation, medications, and procedures. All data were entered into the COPERR database using Research Electronic Data Capture (REDCap) hosted at Atrium Health [20].

We evaluated process-related aspects of CODE PE activations, focusing on time intervals, use of advanced PE interventions, and interhospital transfers. We measured times from PE diagnosis (when radiologist signed computed tomography interpretation) to CODE PE activation and from CODE PE activation to initiation of first heparin dose (unfractionated or low molecular weight). For advanced PE intervention, we reported usage of each type of advanced intervention as well as the time from CODE PE activation to intervention. Advanced PE interventions were defined as systemic thrombolysis (including alteplase 100 mg or 50 mg over 2 h or off-label use of tenecteplase 40–50 mg bolus) and CDI (including ultrasound-assisted catheter-directed thrombolysis [CDT], large and small bore catheter-based embolectomy, and mechanical thrombectomy using devices like AngioVac (AngioDynamics, Latham, NY), Indigo (Penumbra, Alameda, CA), or FlowTriever (Inari Medical, Irvine, CA). We also reported usage of ECMO and surgical embolectomy. Other process outcomes included presence or absence of interhospital transfer, intensive care unit (ICU) admission, ICU length of stay (LOS), and total hospital LOS. For transferred patients, we recorded time from CODE PE activation to discharge from the referring hospital, and time to arrival and admission at the receiving center.

Clinical outcomes included PE-related clinical deterioration and major bleeding during the index hospitalization. Clinical deterioration was defined as cardiac arrest, use of inotropes or vasopressors for symptomatic hypotension (e.g., dobutamine, norepinephrine, dopamine, vasopressin, epinephrine), emergent respiratory interventions, or PE-related death. Respiratory interventions included mechanical ventilation, positive pressure ventilation, or escalation to high-flow oxygen, excluding routine or chronic ventilatory support. Cause of death was based on physician documentation when available; if no alternative cause was listed, death was considered PE-related. For patients presenting with massive PE, continued hemodynamic instability into or during admission was also classified as clinical deterioration. Major bleeding was defined according to the International Society on Thrombosis and Hemostasis criteria [21].

Sample size was based on all patients in the COPERR database with complete data. We calculated counts, percentages, means, and standard deviations for descriptive statistics and reported missing data per variable. Statistical comparisons were made using bivariate analyses: Student’s t tests for continuous variables, and chi-square or Fisher’s exact tests for categorical variables, grouped by whether patients presented to a PE referral or non-referral center (primary dependent variable). For comparisons among the three referral centers (secondary dependent variable), we used ANOVA for continuous and chi-square or Fisher’s exact tests for categorical variables. Analyses were conducted in R and RStudio [22, 23]. A two-tailed p value < 0.05 was considered significant.

For the primary objective, patients were stratified by whether they presented and were diagnosed with PE at a referral center ED or a non-referral center ED. We conducted a subgroup analysis of CODE PE activations at non-referral centers, comparing outcomes between patients who were transferred to a PE referral center vs. those who were not. For the secondary objective, only patients who presented to and were diagnosed at one of the referral centers (Hospitals A, B, or C) were included.

For the exploratory objective, we grouped patients by the date of CODE PE activation into 6-month intervals over the 8-year study period and assessed temporal trends in key variables.

We used a combination of descriptive statistics, visualizations, and inferential models to identify temporal trends. This comprehensive analytic approach was designed to provide insights into how variables changed over time and to identify significant shifts in our data. The descriptive and exploratory analysis involved displaying continuous measurements (e.g., times to treatment) and categorical proportions (e.g., PE severity, advanced intervention use, clinical deterioration, major bleeding) over time.

We used regression to account for the randomness/variability in our raw data to smooth or average out trends and displayed both sets of data. A generalized additive model with normally distributed residuals and a penalized spline was used to model trends over 6-month intervals for each outcome [23]. Predictions were generated on the logit scale and converted to probabilities with 95% confidence intervals using the delta method. Model fit was summarized using estimated degrees of freedom (representing the complexity of the smoothed trend line) and chi-square tests and p-values (representing significant change over time). Degrees of freedom at or close to 1.0 represent linearity. P values less than 0.05 represent significant changes over time. Predicted probability and mean time curves with 95% confidence intervals were plotted to visualize smoothed or averaged changes over time.