Blog

Unpacking Remote Monitoring, EHR Data Security, Migration to CloudHub, Automated Testing, and Fax Automation with AI

Share:

The annual UC Tech conference is coming up this August 9 - 11, 2021. Started in 1982, the conference allows UC faculty and staff to share knowledge across the University of California. The 2021 theme is "Envisioning the Future of IT" and consists of the following tracks:

Register

  • Advancing Research Frontiers with Data

  • Digital Ethics, Privacy & Security

  • Empowering Women and Diversity in the IT Community

  • Envisioning the Future of Healthcare Systems

  • Exploring the Future of 21st Century Learning Spaces

  • Focusing Technology on Teaching and Instruction

  • Infrastructure Across the Spectrum

  • Leveraging Digital Experience (XR) Technologies

  • Professional Development for Technology Professionals

  • Soaring in the Cloud

  • Supporting the University of the Future

  • Universal Design Thinking for Mobile, Web and Wearables

CDHI staff will be sharing their expertise and insights on August 10 and 11. Register here and tune in for these informative sessions.

Building Secure Access to UCSF’s EHR Data

Track: Digital Ethics, Privacy & Security

When: Tuesday, August 10, 2021 9:25 AM - 9:55 AM

Speakers: Sondra Renly, Technical Architect, Christopher Miller, Innovation Engineer, and Veeraragavan Gopalakrishnan, Mulesoft Developer

Description: Electronic Medical Record (EMR) systems contain clinical data critical to driving improved digital patient experiences across our healthcare system. Application Program Interfaces (APIs) provide an industry supported mechanism for securely providing access to our patient data. Current EMR provided security measures through user security settings alone are not sufficient. We will review common API requests, the limitations of Epic's user security settings, and how we have added additional constraints through our MuleSoft proxy to ensure Need to Know access is all that is granted.

The IT Integration Services team has recently transitioned the proxy logic to use dynamic DataWeave scripts and the Salesforce platform to add efficiency and flexibility which has reduced our costs to build and maintain our API data validation and redaction rules. We will explain the initial proxy build, the limitations it had, and how the new dynamic proxy is better serving our needs.

Virtual Care: UCSF's Lung Transplant Chatbot and Remote Monitoring Program

Track: Envisioning the Future of Healthcare Systems

When: Tuesday, August 10, 2021 3:30 PM to 4:30 PM

Speaker: Olivia Bigazzi, Product Manager

Description: The COVID-19 pandemic and resultant need for social distancing accelerated the need to develop remote monitoring of graft function in lung transplant recipients. While home spirometry has been used previously in lung transplantation, prior work shows long-term engagement is poor. We aimed to deploy Bluetooth enabled home spirometers coupled with a digital chatbot to improve engagement and allow efficient data and symptoms.

We designed and implemented an automated, chat-based mobile health intervention paired with Bluetooth-enabled home spirometers. Transplant recipients received text messages or emails, allowing them to engage in a personalized, automated chat with symptom assessment, education modules, and spirometer data collection. The chat engaged patients weekly and allowed them to also initiate chats. Concerning symptoms or FEV1 declines were escalated to the team.

We mailed home spirometers to 424 patients starting April 2020 and launched an automated, interactive chatbot program. Between 5/4/2020 and 10/21/2020, 311 total patients enrolled in the automated chat. Of the 273 patients that submitted at least 1 FEV1 value, they submitted a median 13 (IQR 6-23) FEV1 values over 24 weeks. The largest drop in FEV1 engagement came after the first week in each patient’s chat experience, with 64.7% of those that submitted an FEV1 at baseline continuing to do so into week one and 71.8% in week two. However, after the initial drop, FEV engagement remained stable between 56.3 and 71.8% through 24 weeks.

Lung transplant recipients engaged at high rates with a chat-based mobile health intervention and consistently used Bluetooth-enabled home spirometers to report their FEV1, allowing close home-based monitoring of medically vulnerable transplant recipients. Given the projected need for social distancing, and increasing role of telemedicine for long term management, a chatbot mobile health linked home spirometer may be a powerful tool at detecting early graft dysfunction.

Migrating On-Premises MuleSoft Applications to CloudHub

Track: Soaring in the Cloud

When: Tuesday, August 10, 2021 3:30 PM - 4:00 PM

Speakers: Sondra Renly, Technical Architect, Fel Bautista, Principal Engineer, and Veeraragavan Gopalakrishnan, Mulesoft Developer

Description: Increasingly UCSF is leveraging iPaaS (Integration Platform as a Service) solutions such as Salesforce and ServiceNow. This year we completed a successful migration of our MuleSoft application from on-premises to CloudHub. This project required significant planning, training, and cross-team collaboration. We will discuss change management processes, migration challenges, our introduction of DevOps principles for interface validation, and how we defined outcome metrics.

CDHI DevOps Success Story - Scaling Automated Testing for Connect Care

Track: Infrastructure Across the Spectrum

When: Wednesday, August 11, 2021 11:00 AM - 11:30 AM

Speaker: Sondra Renly, Technical Architect

Description: Our Center for Digital Health Innovation is actively working to improve the new patient experience for those referred to UCSF for our specialty care. Connect Care is a web application we created to greatly reduce the time it takes for referral coordinators to collect common missing information from patients that are blocking us from scheduling appointments. Building our first patient facing web application was challenging for many reasons, but leveraging DevOps and building robust automated testing allowed our team to rapidly identify any newly introduced faults across numerous combinations of desktop and mobile devices to ensure a consistent user experience across every browser and device.

We will share our experience building automated UI, functional, and regression tests in BrowserStack using Selenium and the open source test automation framework Appium. BrowserStack is a cloud based testing infrastructure that gives us live access to real Android, IOS, Mac, and Windows devices with our desired configurations and creates video recordings of each test so that when something is wrong, we can quickly inspect the error log and video log to identify the problem.

Automating Fax Intake with AI

Track: Envisioning the Future of Healthcare Systmems

When: Wednesday, August 11, 2021 2:00 PM - 2:50 PM

Speakers: Ramki Yerramsetty, Product Manager and Bob Rogers, Expert in Residence for AI

Description: Artificial Intelligence (AI) has been integrated into many industry sectors, and healthcare is also undergoing transformation from AI. The Center for Digital Health Innovation (CDHI) at UCSF is using AI to automate manual, repetitive workflows that traditionally have relied heavily on human staff. UCSF receives over 1 million faxes annually. With our Referrals Automation solution, we have been able to use optical character recognition (OCR) technology to automatically extract patient information from faxed referrals. However, this solution is needs further development because the technology is based on using templates of the forms, which both limits accuracy and requires upfront investment to build the templates.

We are working to increase the accuracy, add new automation, reduce upfront investment, and expand the automation to non-referrals, which account for more than 80% of the fax volume. The latest advancements in document image classification and OCR have allowed us to build an application that utilizes these algorithms, automatically and accurately differentiating types of faxes and extracting the the patient information relevant to providers. We are taking incremental steps to streamline the patient intake process, ultimately providing our patients to have a better experience and be scheduled sooner with greater efficiency.

Register