Purpose

The idea for Sofiya came from the issue of nurse burnout syndrome¹. The high workload and chronic stress nurses experience at work can lead to emotional exhaustion, memory loss, impatience and depression ^2,3, which is one of the main reasons nurses leave their job.⁴

One specific task nurses perform daily is the pre-procedure phone call. During this call, the nurse provides patients with instructions for their procedure, verifies important clinical information, and answers any questions the patients may have. This task is highly repetitive, as there are many patients to call.

Every year, the Mount Sinai Cardiac Cath Lab performs thousands of catheterizations. That translates to thousands phone calls, each lasting around fifteen minutes.

Sofiya was designed to relieve nurses of this task, allowing them to focus on more critical responsibilities, such as providing direct patient care in the cath lab.

Of course, Sofiya is an AI and cannot always address every patient concern. A nurse reviews a report generated from Sofiya’s call with the patient and handles any issues that require human intervention. In these cases, the nurse focuses only on specific concerns rather than conducting the entire call. 

Patient Call Workflow

• Sofiya calls the patient and asks them to validate with full name and DOB 
• Intructions are given
• Patient can interupt and ask any questions 
• Questionaire to collect clinical variables
• After each call a report is generated
• NP reviews the report and calls the patient manually if necessary 
• Report is uploaded to Epic

🤖 Agents within Sofiya

Sofiya is powered by a suite of specialized agents designed to manage different aspects of patient interaction ranging from verifying identity and collecting clinical details to answering common questions. These agents work together seamlessly, supported by advanced AI capabilities and conversation memory to ensure personalized, responsive care.

🔐 Guard Rails

To ensure safety, compliance, and reliability, Sofiya is built with robust protections including:

• Intelligent error handling

• End-to-end data privacy and security

• Escalation pathways for unresolved issues

• Voicemail management when live response isn’t possible

​

Sofiya Implementation

Patient Inclusion Criteria:

• Patient scheduled for cardiac catheterization
• Patient is primary contact and no interpreter needed

Every call reviewed by RN for clinical accuracy and patients called back per predefined criteria.

After 90 days of testing, the average nurse interaction time per patient was reduced significantly.

References

1. WHO.  2019. International classification of diseases.
2. Ligibel AJ et al. JAMA Netw Open. 2023;6(12):e2347894
3. Membrive-Jiménez et al. Healthcare. 2022;10(5):954
4. Shah MK et al. JAMA Network Open. 2021;4(2):e2036469