9. Translating Technology Into the Clinic

TL;DR

Machine learning is being increasingly used in clinical decision support systems to provide real-time interventions and suggestions to healthcare providers, but there are still challenges in terms of data availability and system integration.

Transcript

PETER SZOLOVITS: Fortunately, I have a guest today, Dr. Adam Wright, who will be doing an interview-style session and will answer questions for you. This is Adam's bread and butter, exactly how to translate this kind of technology into the clinic. He's currently in the partner system at the Brigham, I guess. But he's about to become a traitor and l... Read More

Key Insights

• 🎰 The adoption of machine learning in clinical decision support systems is driven by the need for more accurate and personalized interventions in healthcare settings.
• 🤩 Data availability, data quality, and integration with existing electronic health record systems are key challenges in implementing machine learning in healthcare.
• 🧑‍⚕️ Standards and frameworks, such as SMART on FHIR, are being developed to facilitate the integration and interoperability of machine learning applications in electronic health record systems.
• 🥳 Third-party vendors play an important role in developing and deploying machine learning models in healthcare settings, working closely with healthcare provider organizations and EHR vendors.

Questions & Answers

Q: What are the main challenges in implementing machine learning in clinical decision support systems?

One of the main challenges is the availability and quality of the data needed to train and validate machine learning models. Integrating machine learning algorithms into existing electronic health record systems can also be challenging due to data privacy concerns and workflow considerations.

Q: How are machine learning models deployed in clinical decision support systems?

Machine learning models can be deployed through various methods, including embedding them directly into the electronic health record system, using standard APIs provided by the EHR vendor, or developing standalone applications that interact with the EHR system.

Q: What is the role of third-party vendors in deploying machine learning models in healthcare settings?

Third-party vendors can play a crucial role in developing and deploying machine learning models in healthcare settings. They can work directly with healthcare provider organizations, either through partnerships or by participating in vendor-specific platforms (e.g., Epic's App Orchard), to integrate their models into existing clinical decision support systems.

Q: How do healthcare organizations deal with cognitive overload caused by excessive alerts from decision support systems?

Cognitive overload is a concern when implementing decision support systems. Healthcare organizations need to carefully design and calibrate the system's thresholds for generating alerts, taking into consideration the workflow and preferences of healthcare providers. Additionally, periodic evaluation and feedback from healthcare providers can help fine-tune and refine the system to reduce unnecessary alerts.

Summary & Key Takeaways

• The adoption of AI-powered technologies in healthcare is following a typical hype cycle, with initial excitement followed by a decline in expectations and eventual acceptance and adoption.

• Clinical decision support systems have been traditionally rule-based, providing alerts and reminders to healthcare providers. However, there is a growing interest in using machine learning algorithms to provide more accurate and personalized recommendations.

• Despite the progress, there are still challenges in terms of data availability and integration with existing electronic health record systems. Data quality, data privacy, and workflow considerations are key factors that need to be addressed.

• There are ongoing efforts to develop standards and frameworks, such as SMART on FHIR, to facilitate the integration and interoperability of machine learning applications within electronic health record systems.