NB: this series is still a work in progress.
The Healthcare AI Lifecycle
Artificial Intelligence (AI) holds great promise for advancing the practice of medicine. Modern evidence-based medicine (EBM) practice depends on synthesizing copious amounts of information across large populations of patients. AI’s subdomain of Machine learning (ML) provides a set of techniques to build data-driven prediction models to fulfill the goals of EBM. Complex medical information synthesis tasks, such as detecting patients at risk for uncommon conditions, may be aided by using AI models. Over 500 Food and Drug Administration (FDA) certified AI systems support physicians in various tasks, ranging from electrocardiogram analysis to mammogram breast cancer detection. Health systems and health information technology (HIT) vendors are also developing and implementing AI systems that do not require FDA certification. These systems are meant to inform physicians by providing risk estimates that can be incorporated into medical decision-making.
These systems are a prime example of healthcare AI, tools that are used by patient, clinicians, and health systems to improve or inform health, clinical decision making, or operations. As alluded to above, there are many different types of healthcare AI systems. Despite different use cases these systems all share similarities in how they are made and used. This is because there are certain things that will need to be done in order to ensure that these systems are safe and effective for use in healthcare. I refer to the all of these steps as the healthcare AI lifecycle.
Understanding the Healthcare AI Lifecycle
There are two major phases to the lifecycle:
- Development - making AI tools (known as models)
- Implementation - using the AI tools I would also argue that like all software development, making healthcare AI is a continuous journey, so it doesn’t end once a model is in use.
Overview of Development
Development are the processes involved in creating an AI model. One of the first issues is access to datasets. Having obtained data, model developers may realize that healthcare data, like healthcare itself, is complicated. Processing and transforming data for AI model development requires a unique mix of clinical and technical expertise. After being developed, models must be internally and externally validated (in some settings) to assess if they will benefit patients, physicians, or healthcare systems. However, external validation may be challenging due to data-sharing restrictions.
Overview of Implementation
Implementation is the work of integrating and utilizing an AI model into clinical care. The implementation process may begin once a model is validated. Implementation raises a host of issues that are rarely confronted during model development. The technical work needed to implement models often requires cobbling together disparate HIT systems, such as databases, web services, and electronic health record (EHR) interfaces. Additionally, implementation requires special attention to human factors and systems design. These models are not used in a vacuum; developers must carefully consider users and their workflows. Finally, there is the issue of monitoring and maintaining these AI systems. As healthcare systems change, AI systems may experience performance degradation due to patient populations or medical practice changes. Thus, developers may need to update their models over time. Despite their promise, successfully developing, implementing, and periodically updating AI models for healthcare is a challenging engineering task.
Where to go from here?
This post is meant to be an introduction. Its pretty general so that we can use it as a spring board to explore all the interesting aspects of using, making, and breaking healthcare AI systems.
This post was co-developed with a couple related posts, two covering development and implementation in more detail and a third discussing the HIT infrastructure needed to support making and using healthcare AI tools.
Some of this content was adapted from the introductory chapter of my doctoral thesis, Machine Learning for Healthcare: Model Development and Implementation in Longitudinal Settings.
Cheers,
Erkin
Go ÖN Home
