🚨 Pulse of Innovation 🚨

Breaking news in the healthcare AI

Coalition for Health Artificial Intelligence - CHAI

The goal of the CHAI “is to develop “guidelines and guardrails” to drive high-quality health care by promoting the adoption of credible, fair and transparent health AI systems.”

• Twenty non-profit hospital centers across US including Duke, Johns Hopkins, Kaiser, Vanderbilt and other along with industry partners Amazon, Google, Microsoft and CVS health came together and formed the coalition to provide framework and guidelines to fair, transparent and trustworthy adoption of Health AI systems.

• The leadership includes physicians, data scientists, technology partners and federal government liaisons to ensure safe and more regulated approach towards developing the Health AI guidelines and guardrails.

• The CHAI provided the blueprint for trustworthy AI implementation guidance and assurance for healthcare in March 2024. This covers wide range topics including reliability, resilience, real-world utility, systemic bias, regulatory compliance, model testability, transparency and interpretability and many more.

• The CHAI has now announced the framework for responsible development and deployment of artificial intelligence tools in healthcare and has it open for 60 days for public comment.

• The CHAI initiative is designed at building and enhancing trust in Healthcare AI for its stakeholders who are clinicians, administrators and above all patients. Although CHAI consist of mostly larger academic health systems its open for anyone to enroll and be part of the process to make the health AI secure and trustworthy.

  

🧑🏼‍🔬 Bench to Bedside👨🏽‍🔬

Developments in healthcare AI research and innovations

Utilizing GPT-4 for Enhanced Clinical Trial Screening: A Game Changer

• To determine patient eligibility for the COPILOT-HF study beyond what structured EHR queries can identify, researchers developed the RAG-Enabled Clinical Trial Infrastructure for the Inclusion Exclusion Review (RECTIFIER).This innovative system leverages a clinical note-based, question-answering approach powered by Retrieval-Augmented Generation (RAG) and GPT-4.

• Clinical notes from 100 patients were used for the development dataset, 282 for the validation dataset and 1894 for the test dataset.An expert clinician conducted a blinded review to establish gold standard answers for 13 target eligibility criteria questions.Researchers calculated performance metrics including sensitivity, specificity, accuracy and the MCC to evaluate patient eligibility.

 Key findings:

• Accuracy and efficiency: The AI model demonstrated remarkable accuracy, ranging from 97.9% to 100%, in aligning with expert clinician assessments for trial eligibility. In contrast, traditional screening performed by staff showed slightly lower accuracy rates between 91.7% and 100%. 

• Cost Effectiveness : Cost of using this AI model is significantly lower, estimated at approximately $0.11 per patient, compared to traditional methods.

• Potential for Broad Applications: The study highlights the potential of integrating AI into clinical trial workflows to expedite the trial process, reduce costs, and enhance participant diversity.

• Challenges and Safeguards: The researchers emphasize the need for ongoing monitoring to prevent biases, missing out the clinical nuances and ensure the system’s adaptability to changes in data capture methods.

  Conclusion:

• The adoption of GPT-4 in clinical trial screening marks a significant advancement in medical research methodologies. By enhancing the accuracy, efficiency, and cost-effectiveness of participant selection, this AI-driven approach holds the potential to transform clinical trials, making them more accessible, representative and faster.

  

  Demonstration of the RECTIFIER WorkFlow ( Source: NEJM AI)

👩‍⚕️AI in the Clinic 🏥

Real-world and practical use of healthcare AI in clinic

AI powered tools for control of Myopia in children: A story from far east

• Early onset myopia in children age 3-18 years is on rise and the WHO estimates that by 2050 over 70% of children in this age group will be myopic in Asia.

• Long standing myopia can be detrimental to vision and early detection and treatment is key in preventing this sequalae and preserving the vision.

• With this goal the FV hospital in Vietnam has launched a program for early detection of myopia in children age 3-18 years using AI powered tools.

• The hospital is using Myopia Master Measurement which can take measurements of the eye and can accurately predict patients in pre-myopia, at high-risk of developing myopia, and timing around which they are most likely to develop myopia. The measurements use AI for this accurate prediction.

• Once the patients are identified at higher risk the hospital implementing myopia control protocols and frequent measurements to ensure progression and effectiveness of control regimen. Widespread implementation of this could significantly reduce the occurrence of myopia and control its progression in children.

  

🩺 Start-Up Stethoscope 💵 

Trending startups and technologies impacting clinical practice

Aiding senior care with AI: Better detection and action on care anomalies. 

• Senior care is often complex and involves multiple avenues that includes medication dose adjustments, other care management and physician led clinical decision making to ensure compliance to the care prescribed. Due to multiple comorbid conditions care management may need frequent changes and adjustments.

• Sensi.ai uses specialized audio technology that records and analyzes senior care information at various time points and uses AI to detect care deficiencies, necessities and notifies the care provider immediately. 

• Sensi.ai has a HIPAA compliant plug-in audio pods that record the patient care conversation with precision and transmits to the team for further evaluations. Sensi provides insights on caregiver-client dynamics, enabling coordinated responses, caregiver rewards, and optimal pairings.

• Sensi identifies care patterns and provides insights, spotting trends for early action. It is easy to use and can be integrated to different care platforms. A valuable addition for senior care enabled by AI. 

• It has recently raised $31 million in series funding with anticipation of growing senior care market with widespread AI adaptations.

🤖Patients First, AI Second🤖

Ethical and regulatory landscape of healthcare AI

FDA's EDSTP: Advancing AI and Machine Learning in Pharmacovigilance

• The FDA's Emerging Drug Safety Technology Program (EDSTP) is an initiative under the Center for Drug Evaluation and Research (CDER). This program is designed to leverage advanced technologies, including machine learning in drug development and safety.

• The EDSTP focuses on adopting and integrating new technological advancements to better detect, assess, understand, and prevent adverse effects or other drug-related problems by administering the Emerging Drug Safety Technology Meeting (EDSTM) Program , announced on 11th of June, 2024.

• It addresses key challenges in pharmacovigilance, such as efficient data collection, processing, and evaluation, using emerging technologies like AI.By integrating AI to automate data collection and compilation, the program aims to significantly reduce administrative burdens and costs.

• This initiative encourages communication between the pharmaceutical industry (including academia, contract research organisations, pharmacovigilance vendors, and software developers) and the FDA. The goal is to discuss the research and development of AI and other emerging technologies in pharmacovigilance and potential policy and regulations for such tools.

  

   Read here for more information on EDSTMP