The following is a guest article by Alex Flores, General Manager, Health and Life Sciences Vertical at Intel, and Burnie Legette, Senior Solution Architect at Intel
Healthcare practitioners rank among the world’s most valuable employees, but right now they need help, too. Nearly half of physicians feel burnt out, according to the American Medical Association.
Endless administrative work, along with other pressures, fuels their stress. For instance, it’s not unusual for a primary care physician to spend two hours of electronic health record documentation for a one-hour patient encounter. The administrative burden can be even higher in oncology, where the EHR usage is contributing to physician burnout.
AI isn’t the cure-all for the burnout challenge, but it can streamline workflows, speed up scheduling and transcription, and help doctors deliver more accurate, data-driven recommendations at the point of care. Yet tight budgets, staffing shortages, and outdated systems continue to slow modernization and limit AI’s impact in hospitals and physician practices.
Fortunately, there are steps healthcare organizations can take to implement AI in small, secure, but highly impactful ways. By mapping workflows to strategically target AI deployments, leveraging interface engines and APIs, and adopting federated data practices, healthcare IT teams can successfully enhance their systems and make life easier and better for both physicians and patients.
Workflow Mapping Helps Uncover AI Needs
Workflow mapping provides a structured way for health leaders to uncover where AI can add the most value without changing healthcare infrastructure. With careful analysis, IT teams can pinpoint areas where inefficiencies, bottlenecks, or repetitive tasks that routinely consume staff time.
They can start by conducting department-level assessments that break daily operations into clear, step-by-step processes. Then work with both clinical and administrative staff to document where delays, redundancies, or manual workarounds occur. Feedback, time studies, and system logs reveal where people get bogged down the most.
This process helps IT leaders determine where AI will be most impactful. Routine responsibilities, such as appointment scheduling, data entry, prior authorization requests, and transcription, are all good places to start. AI can automate patient scheduling and check-ins through chatbots, intelligently process documentation for prior authorization, and perform other tasks.
AI can also enhance patient interactions and care by providing actionable intelligence from image scans, medical histories, and other data, saving doctors time and allowing them to focus on delivering personalized treatment and spending more meaningful time with patients.
APIs and Interface Engines Make AI Easy to Implement
APIs and interface engines (also known as integration engines) make it possible to achieve these outcomes with minimal IT changes. New AI applications can connect with existing infrastructure through APIs, while interface engines transfer data more easily across disparate healthcare devices and systems.
This approach allows organizations to add valuable innovative AI features without redesigning user interfaces or hardware. For example, sonographers often juggle a transducer in one hand while typing commands with the other, which can be challenging and error-prone. Multi-modal AI interfaces enable sonographers to easily switch modes on the ultrasound machine by speaking commands.
Data science teams can also optimize AI algorithms to run on existing edge devices and IT architectures. For instance, AI models deployed on an MRI machine can be used to correctly position patients before a scan, or can help a sonographer produce higher-quality images, leading to a more accurate diagnosis. Running AI on established technology and at the point of care keeps costs down and provides physicians with real-time data they can use to guide treatment.
APIs and interface engines enable modular software packages that can be written once and reused across different devices and modalities. The results can lead to faster integration and minimal disruption to daily workflows.
Federated Data Keeps Patient Information Secure
Public AI models train on large collections of publicly available data. However, this approach isn’t feasible for healthcare organizations, which deal with highly sensitive patient records and tight data privacy regulations.
Federated learning strategies allow organizations to share data across institutions and geographies without exposing or transferring protected health information. This training method enables healthcare organizations to keep patient data secure in-house while still improving the accuracy of AI models. Instead of transferring sensitive information, systems send the model to the data, train it locally, and then share the updated insights back. This closed-loop approach protects patient privacy while building intelligence from data spread across different devices and clinical practices within the network.
There are several noteworthy examples of federated learning’s efficacy. In particular, the National Cancer Institute’s federated learning network via the Center for Biomedical Informatics and Information Technology shares information between leading universities and cancer centers. Similarly, the Federated Tumor Segmentation consortium marshals data from 71 sites across six continents to perform federated glioblastoma tumor segmentation, achieving up to 33% improvement in tumor margin delineation versus centralized approaches.
Conclusion: AI as an Augmenter, Not a Disruptor
AI is a disruptive technology, but that doesn’t mean it will disrupt hospital IT infrastructure. Instead, it should complement existing IT resources and augment the capabilities of a healthcare organization, right down to the point of care.
By identifying areas where AI can provide the most value, leveraging APIs and interface engines, and employing federated learning, providers can efficiently integrate the technology into their existing processes with minimal impact on IT. Implementing AI will reduce clinicians’ administrative burden, alleviate some of their stress, and elevate their ability to provide personalized, accurate, high-quality care.
No comments:
Post a Comment