A patient interaction turned into clinician notes in seconds, increasing patient engagement and clinical efficiency.
Novel compounds designed with desired properties, accelerating drug discovery.
Realistic synthetic data created at scale, expediting research in rare under-addressed disease areas.
These are just a few examples of how generative AI and large language models (LLMs) are transforming the healthcare and life sciences (HCLS) industry. Generative AI uses neural networks and deep learning algorithms from LLMs to identify patterns in existing data to generate original content. But while the potential is theoretically limitless, there are a number of data challenges and risks HCLS executives need to be aware of when using AI that can create new content. Accessibility, quality and security are crucial—according to McKinsey, “Your data and its underlying foundations are the determining factors to what’s possible with generative AI.”
Here’s how the right data strategy can help you get past the hazards and hurdles to implementing gen AI.
Generative AI applications in HCLS
According to a recent KPMG survey, 65% of U.S. executives believe generative AI will have a big impact on their organization in the next three to five years. That’s because gen AI has many use cases across the enterprise and particularly for healthcare providers, payers and life sciences organizations. Here are some examples for each subsector.
- Healthcare providers: Gen AI is revolutionizing clinical practice and patient care delivery. It can use natural language processing (NLP) to automate the process of medical documentation, significantly reducing the administrative burden on healthcare workers and allowing them to focus more on patient care. Gen AI can also analyze unstructured data sets, such as clinical notes, diagnostic imaging and recordings and provide evidence-based recommendations. Doctors are using gen AI to create personalized patient communications and treatment plans, improving the patient experience and ultimately creating better health outcomes.
- Healthcare payers: Gen AI is accelerating operational efficiencies and providing more targeted evaluations for healthcare payers. It can use LLMs to mine clinical and environmental data and provide more accurate risk assessment scoring. In the absence of real data, it can create synthetic data that can be used to improve the performance of predictive models to assess risk. It can detect pre-payment fraud, waste and abuse (FWA) and ensure provider integrity through data accuracy scoring and auto-credentialing. It can also increase agent productivity through chatbots that can handle basic customer queries, allowing agents to focus on more complex cases.
- Life sciences: Gen AI is fundamentally altering the life sciences landscape. It is being used to design novel compounds with specific functionalities, design synthetic gene sequences for applications in biology and generate synthetic data to augment existing data sets for improved AI model training and performance. Companies are using generative models to create synthetic patient and healthcare data to simulate real-world data, expediting drug discovery and enabling them to better understand patient populations and diseases.
Pitfalls and perils
The success of these use cases is highly dependent on the data used to train generative AI. As HCLS executives integrate generative models into their analytics and AI roadmaps, they need to be aware of the issues associated with the data that feeds the models.
- Data quality and access: Access to high-quality data has always been a vital element of AI, but it becomes even more important when dealing with the scale and scope of data that gen AI models rely on. After all, training one LLM can cost millions of dollars. In addition, healthcare data is frequently incomplete, unstructured and stuck in data silos. An inconsistent data set introduces biases and inaccuracies, which can have profound consequences for clinicians or scientists using an AI model for patient health.
- Data security: In a recent KPMG survey, 63% of business leaders surveyed rated privacy concerns with personal data as a top risk associated with gen AI, while 62% rated cybersecurity as a top risk. Here’s an example of how data security may be a risk: During training, a foundation model might have access to all the data available within the organization, including individuals’ personally identifiable information (PII) and sensitive corporate data. If the model uses ungoverned data to generate content, there is a risk of exposing sensitive data to unauthorized data consumers.
- Regulatory compliance: The healthcare industry is subject to some of the most stringent data privacy regulations around the world. The United States’ Health Insurance Portability and Accountability Act (HIPAA) and the European Union’s General Data Protection Regulation (GDPR), for example, set strict standards for protecting health information. In the KPMG survey, respondents ranked concerns about the regulatory landscape as the biggest barrier to adopting gen AI. Organizations can’t rely on automatically generated code for system implementations or mandatory reporting without safeguarding that it is consistent with industry and government requirements.
- Computing resources: Training a foundation model involves harnessing and processing a massive amount of structured and unstructured data, which requires high processing power and storage space. These types of resources are often costly to purchase and manage. In addition, hiring for AI-related roles such as AI data scientists, data engineers and AI product owners remains a challenge.
A plan of action
To overcome these risks and challenges, HCLS executives need to adopt a generative AI data strategy that includes three key elements:
- The ability to collect, process and analyze data in one location: Good data hygiene is essential. When the structured and unstructured data is in one place, it’s easier to develop and enforce firm quality control standards and establish validation protocols for AI-generated content.
- Strong governance and security: Robust governance and security practices help ensure data is protected from unauthorized access and that the company is able to remain in compliance with data protection regulations. They also ensure the integrity of the AI models themselves, preventing malicious agents from tampering with their outputs.
- A flexible, scalable, managed data infrastructure: A cloud-based data infrastructure offers capacity without needing to purchase or update equipment. It should be able to scale up and down on demand, ensuring that additional users can be accommodated without competing for resources. It should also be able to perform optimally on its own without constant manual oversight.
The solution: A modern data cloud platform
A modern data cloud platform offers the capabilities needed to access, process and manage high-quality data and keep it secure. Here are some of the capabilities of a modern data cloud platform:
- A single repository for data, breaking down data silos and simplifying data management
- Secure and seamless data sharing across systems, organizations and clouds
- A high level of data security, enabling customers to comply with evolving regulations
- Continuous optimization of price for performance so customers only pay for what they use
- Flexibility and scalability to process massive amounts of data from different sources without resource contention
- The ability to scale the development, operationalization and consumption of ML models across the enterprise
- The ability to easily build, distribute and monetize applications
Discover how Snowflake’s Healthcare and Life Sciences Data Cloud can help you unlock the potential of generative AI.