This study evaluates the impact of a quality improvement project designed to increase the rate and timeliness of GOC conversations for patients at high risk for mortality in the CICU via a retrospective, observational, quasi-experimental approach.

Approval for the study was obtained by our local institutional review board. The study satisfied the requirements for waiver of consent.

Our project was conducted in the CICU and Cardiac Stepdown Unit (CSDU) at The Mount Sinai Hospital, a large, tertiary academic medical center in New York City. The units have 14 beds and 6 beds, respectively, and specialize in the comprehensive management of patients with complex arrhythmias, advanced heart failure, mechanical circulatory support, and post-cardiac transplantation care. The team includes an attending cardiologist, an attending intensivist, medical trainees (fellows and residents), specialized nurses, respiratory therapists, speech therapists, nutritionists, physical therapists, and a unit-based social worker. Specialty palliative care is available on a consultative basis.

In the CICU and CSDU, GOC conversations were historically initiated at the discretion of the primary attending clinician in collaboration with the unit social worker. These efforts frequently emanated from informal discussions regarding patients’ palliative care needs at daily multidisciplinary rounds; however, no formal structures were in place to identify patients requiring GOC conversations or to initiate them.

Our intervention was conducted between December 2021 and September 2022. It leveraged a validated tool for mortality prediction, the Surprise question (SQ), which asks the primary clinician: “Would you be surprised if this patient died within the next 6 months?” The framework for our intervention was inspired by NYU’s Supportive Care Program, but adapted to our unit’s unique workflows and institution’s EHR infrastructure, and the novel setting of the CICU [16, 17].

Implementation of the SQ-based workflow proceeded in two phases. During the feasibility pilot (December 2021-March 2022), the primary clinician answered the SQ using an Excel spreadsheet for all CICU patients upon admission, without changing other workflows. The unit’s informal approach to coordinating GOC conversations continued as outlined above. Therefore, this cohort represents the CICU’s baseline for conducting GOC conversations, and is referred to as the pre-intervention group.

Our hospital utilizes Epic (Madison, WI) for its EHR system. Immediately following the conclusion of the feasibility pilot in March 2022, a new column was built and integrated into the attending patient list dashboard. On admission to the CICU, a red icon would appear within the SQ column on the patient list. This icon served as a non-intrusive alert for the attending to answer the SQ for that particular patient. Double-clicking on the icon would bring up a pop-up window with the SQ. If the attending selected “Yes,” no further action was required beyond standard care, and the red icon was removed. If the attending selected “No,” the red icon was replaced with an orange icon that served as a non-interruptive notification to have a goals of care conversation. To turn the icon green, the attending has to complete and document a family meeting using a note template designed for the project. This template prompted the clinician to note the key elements of the conversation, including the shared plan of care. Once the conversation and note were completed, the orange icon was replaced by a green icon, indicating that the workflow was completed (see Fig. 1; see Supplementary Figs. 1–3). Of note, while other clinicians, such as trainees or consultants, may have participated in or had separate discussions, only the CICU attendings were able to document the GOC conversation as part of the official workflow. Our study tracked these CICU attending-led conversations.

Intervention workflow to promote primary palliative care delivery in the CICU. On all CICU admissions, a red alert populates the patient-list dashboard, requiring the attending clinician to answer the SQ via a pop-up window. 1. If the clinician selects “Yes,” no further action is required beyond standard care. The red alert is automatically cleared and replaced with a green icon. 2. If the clinician selects “No,” a non-interruptive orange alert is activated, indicating a goals-of-care conversation is required. 3. To clear the orange alert, the clinician must complete and document a family meeting using a smartphrase template, which prompts the clinician to note the key elements of the conversation including the shared plan of care. Once complete, the orange alert is automatically cleared and replaced by a green icon

All patients admitted to the cardiac intensive care unit during the study period with SQ = “No” were eligible. No patients were excluded.

A list of all patients with SQ = “No” during the pre-intervention period was obtained via Excel spreadsheet. Post-intervention, all patients with SQ = “No” were identified via EHR query. Pre- and post-intervention groups were compared based on demographics, CICU length of stay, rates of ICU interventions (mechanical ventilation, continuous veno-venous hemofiltration (CVVH), hemodialysis, CHF solution), code events and mortality in the CICU. These clinical data were also obtained via EHR query.

Primary outcomes included: (1) the presence, and (2) the timeliness of goals-of-care conversation documentation. For the pre-intervention group, two readers reviewed all notes documented in Epic and counted patient-care team interactions as GOC conversations if the note addressed either: (1) prognosis and/or illness understanding or (2) goals and/or treatment options. These criteria were chosen based on existing literature and the clinical expertise of the research team as core components of shared decision-making [18]. Any discrepancies between the two readers’ assessments were resolved by the broader author group. Usage of the newly created ACP note template that attendings were directed to use was also tracked. We manually reviewed a random selection of these notes to ensure they met the same criteria used to evaluate the pre-intervention notes.

Secondary outcomes included a change in code status, discharge to hospice care, or transfer to the palliative care unit. Together these measures represent GOC redirected towards comfort and non-escalation. These data were collected via EHR query.

Descriptive statistics summarized demographic and clinical characteristics. Continuous variables were presented as medians and interquartile ranges (IQR) and are compared between groups using the Mann-Whitney U test. Categorical variables were reported as frequencies with percentages and are compared between groups with the Chi-square or Fisher exact test as appropriate. To minimize the potential for Type I error, we selected an alpha level of 0.01 for statistical significance. No formal sample size calculation was performed, because the study was conducted as part of a quality improvement project in a real-world clinical setting, where controlling the sample size was inherently challenging. Analyses were performed using SPSS (IBM, Armonk, NY, USA).