Digital health with Dr. Oliver Aalami

Key Concepts:

• Biodesign for Digital Health: A framework for healthcare innovation starting with a well-defined need.
• Needs Finding: The process of identifying unmet needs in healthcare by observing and interviewing stakeholders.
• Needs Statement: A structured description of a problem, the affected population, and a measurable outcome.
• Needs Criteria: Specifications and requirements for a potential solution, including efficacy, cost, safety, and usability.
• Spezi: An open-source framework designed to accelerate mobile health research and prototyping.
• HL7 FHIR: A data standard for healthcare interoperability.
• 21st Century Cures Act: Legislation mandating that certified electronic health records expose APIs for patient access.
• Integrated Delivery Systems: Healthcare organizations that combine payer, provider, and pharmacy functions.
• Self-Insured Employers: Companies that directly pay for their employees' healthcare.
• Medicare Advantage: A Medicare program administered by private companies.
• Translation: The process of converting AI tools and digital health potential into real-world applications.

1. Dr. Oliver Aalami's Background and Inspiration

• Dr. Aalami's early exposure to medicine through his grandparents, who were family physicians in Germany, significantly influenced his career path. He observed their tight connection with the community and was inspired by their impact.
• He studied molecular biology at UC San Diego, attended Boston University for medicine, and completed a general surgery residency at UCSF.
• He spent two years at Stanford for a postdoc, where he became involved with the biodesign program.
• He completed a vascular surgery fellowship at Northwestern in Chicago, specializing in treating arterial and venous conditions.
• He chose surgery because he is a "doer" and wanted to see immediate impact, such as saving lives in critical situations.
• He is passionate about translating AI tools and digital health potential into real-world applications.

2. Biodesign for Digital Health Course

• The Biodesign program was started over 20 years ago by interventional cardiologists who innovated on medical devices.
• The Biodesign for Digital Health course was started in 2012 by Marta Zanchi, an electrical engineer.
• The course uses a design-thinking process, starting with identifying a well-defined need in healthcare.
• The process involves being a "fly on the wall" in clinical settings or following individuals with chronic conditions to understand their needs.
• The goal is to de-risk a concept before investing significant resources.
• The Zio Patch, a wearable Holter monitor, is cited as a successful example of a company that came out of the Biodesign program.
• The course is solution-agnostic, focusing on needs finding before considering digital solutions.
• Needs finding now extends beyond the hospital to include the patient's home and daily life, leveraging connected devices.

3. Needs Finding Process

• Students are assigned a theme, such as pain, and tasked with identifying problems that need to be solved.
• Multidisciplinary teams, including engineers, social scientists, physicians, and policy experts, collaborate to examine problems from all angles.
• Students use platforms like Reddit and Facebook to engage with people and gather insights.
• The needs-finding process is all-encompassing, considering all aspects of a patient's experience, including their activities, diet, and interactions with the medical system.
• For each unique insight, a needs statement is created, defining the problem, the affected population, and a measurable outcome.
• The population can be defined narrowly (e.g., patients after an ankle fracture with swelling) or broadly (e.g., patients with ankle pain).
• The measurable outcome must be achievable within a reasonable timeframe.
• The fellows generate almost 300 need statements, which are then filtered down through research and stakeholder feedback.

4. Stakeholder Considerations

• The needs-finding process includes all stakeholders, such as insurance companies, billing personnel, and surgeons.
• The best concepts address the needs of all stakeholders to increase the likelihood of adoption.
• Healthcare is more complex than the consumer world because patients do not directly pay for their care.
• The insurance company pays the doctor, creating a convoluted system.

5. Steps After Needs Finding

• After needs finding, students create needs statements and needs criteria.
• Needs criteria define the product requirements, including efficacy, cost, safety, and usability.
• The next step is brainstorming concepts.
• The concepts are weighted based on how well they meet the needs criteria and needs statement.
• Early prototyping is crucial for testing hypotheses.

6. Spezi Framework

• Spezi is a prototyping tool designed to accelerate the development of mobile health applications.
• It evolved from Cardinal Kit, a template application.
• Spezi is modular, with components for Bluetooth connectivity, onboarding, surveys, and active tasks.
• It uses Google Firebase for authentication and Firestore as the database.
• The Spezi data pipeline, a Python library, helps researchers quickly gain insights from their data.
• Spezi is completely open source and available on GitHub.
• The rationale behind open-sourcing it was to build and grow the ecosystem.
• Spezi complies with the HL7 FHIR data standard for interoperability.

7. AI Integration

• AI was integrated to analyze the large amounts of time series and high-frequency data collected from patients.
• The availability of tools like TensorFlow and research in computer vision at Stanford facilitated AI integration.
• One application is using AI to pre-autofill medical history forms by leveraging data from the patient's phone.
• The 21st Century Cures Act mandates that certified electronic health records expose APIs for patient access to their health data.
• Patients can use their login credentials to pull their data onto their phones and then use LLMs to interact with the data.
• A demo called HealthGPT allows users to have a dialogue with their health data.

8. Large Language Models (LLMs) and Healthcare

• LLMs are being used to interact with health data on a conversational level.
• A study is being conducted using an application called LLM on FHIR to allow patients to ask questions about their hospital stay.
• LLMs can help patients understand complex medical information and reduce anxiety.
• Foundation models for images can help patients understand X-rays and other medical images.
• LLMs can translate medical information and explain it at a fifth-grade level.

9. Program Application and Target Adopters

• The Biodesign fellowship is open to applicants, while the courses are for Stanford students.
• Policy fellowships are also available.
• Integrated delivery systems, self-insured employers, and value-based alternative payment models are early adopters of digital health solutions.
• Integrated delivery systems like Kaiser Permanente are incentivized to connect with patients in low-cost settings for prevention and prediction.
• Self-insured employers like Google are interested in digital solutions to manage employee healthcare costs.
• Medicare Advantage plans are incentivized to keep patients healthy to avoid high costs.

10. Balancing Multiple Responsibilities

• Dr. Aalami balances clinical work, research, teaching, and innovation.
• His academic position allows him to dedicate time to both clinical practice and research.
• He is passionate about innovating and improving healthcare processes.
• He sees the potential of AI tools to scale access to healthcare.

11. The Evolving Definition of a Person of AI

• The definition of a person of AI has expanded from researchers and developers to include consumers.
• Dr. Aalami considers himself a person of AI because AI is integrated into his course, work, and app development.
• He believes that the key is translation, taking AI innovations and applying them to real-world problems.
• The next generation of physicians will have a background in development and be able to implement AI models to solve problems they encounter in their training.

12. Conclusion

Dr. Aalami's work at Stanford Biodesign exemplifies the intersection of medicine, technology, and innovation. By focusing on needs finding, stakeholder engagement, and open-source development, he is helping to create digital health solutions that can improve patient care and access to healthcare. The integration of AI, particularly LLMs, offers new opportunities for patients to interact with their health data and better understand their conditions. His emphasis on translation ensures that these innovations are not just theoretical but have practical applications in the real world.