Meet the all-new ClinicalKey AI, with expanded full-text content and upgraded features

ClinicalKey AI is a clinical decision support tool that combines trusted, evidence-based content with conversational search powered by generative AI. It is designed to support clinicians in making informed decisions at the point of care by quickly finding and summarizing relevant source content to deliver reliable responses to their queries.

ClinicalKey AI uses a complex RAG (Retrieval Augmented Generation) system that augments multiple LLMs (large language models) with search accuracy of our vectored content database. This system retrieves the most relevant documents using a vector search technique that finds semantically similar content rather than relying on simple keyword matching. An LLM then composes the information in these documents into a concise summary optimized for supporting clinical decision-making.

This multi-step approach, coupled with our clinician-in-the-loop AI evaluation process, helps minimize hallucinations and bias common in traditional LLMs and produces precise, contextually appropriate responses to a user's query.

ClinicalKey AI draws exclusively from trusted, copyright-cleared proprietary and third-party sources, ensuring access to reliable clinical evidence. Each response includes citations linked to published evidence, promoting trust and enabling further exploration. 

Content sources include:

• Clinical guidelines from leading medical societies, such as ACC, AACE, ACEP, CHEST, EMSO and others

• High-impact Elsevier and non-Elsevier journals, such as The Lancet, Clinics Review Articles, New England Journal of Medicine and Journal of the American Medical Association

• Medical reference books including Braunwald’s Heart Disease, Nelson Textbook of Pediatrics and more

• Elsevier’s proprietary ClinicalKey point-of-care content

• Drug information and more

ClinicalKey AI mitigates risk and fosters trust by using a secure platform of copyright cleared content, including proprietary Elsevier content and trusted third-party sources. This provides breadth and depth in the information available to clinicians. 

Yes, each summary generated by ClinicalKey AI includes citations that link directly to the supporting published evidence. This allows clinicians to explore further when needed and verify the information. 

Additionally, real-time citation validation ensures each datapoint and assertion in the generated response is checked against cited references for accuracy.

References are selected primarily for relevance. When available, the system prioritizes evidence types that commonly form the core of clinical decision-making—such as practice guidelines, meta-analyses, Cochrane reviews, government or expert panels, and large randomized controlled trials.

For over 10 years, Elsevier has used AI and machine learning responsibly in our products, combining it with unparalleled peer-reviewed content, extensive data sets and sophisticated analytics. 

ClinicalKey AI is developed in line with Elsevier's Five Responsible AI Principles: 

1. We consider the real-world impact of our solutions on people. 

2. We take action to prevent the creation or reinforcement of unfair bias. 

3. We can explain how our solutions work. 

4. We create accountability through human oversight. 

5. We respect privacy and champion robust data governance. 

Elsevier also aligns to industry-standard guideance from organizations like Coalition for Health AI (CHAI) and the College of Healthcare Information Management Executives (CHIME).

We employ a clinician-in-the-loop evaluation framework, helping to ensure ClinicalKey AI meets the high standards required for clinical use. Quarterly and before major releases, we work with licensed clinicians across major medical specialties to review query-directed responses to assess the tool for accuracy and bias. As a member of CHAI, our evaluation team has contributed to CHAI’s testing and evaluation framework guidelines.

ClinicalKey AI is built with a security-first philosophy and is built to be HIPAA-compliant. ClinicalKey AI encrypts all user data in transit and at rest using industry standard protocols. Access is strictly limited to authorized personnel under pseudonymization for product support only, with no access granted to cloud providers or third-party AI partners for training purposes. Rigorous security vetting, assessments, and continuous vulnerability scanning further protect user data.