Ranjani Ramamurthy, Principal Physician Scientist
The concept of ambient intelligence has been around since the late 1990’s, but has only really taken off in recent years (partly due to the success of consumer technologies, like voice-activated smart home assistants).
An “ambient” intelligence is a system that works invisibly in the background, largely unobtrusive, until it can weigh in with useful contextual actions or information. The best ambient intelligences are ubiquitous, pervasive, and human-centric.
At Project EmpowerMD, a Microsoft incubation project, we are guided by the fourth pillar of the Quadruple Aim in healthcare, which states that healthcare outcomes can be improved by taking better care of doctors.
The last decade has seen large-scale adoption of Electronic Health Record (EHR) systems. The EHR has undoubtedly had many benefits, but it has dramatically changed the nature of a doctor’s work and resulted in some unintended consequences. Doctors nowadays spend approximately 60% of their time on electronic paperwork, effectively tethered to the keyboard. Most of this time is spent on clinical documentation and information retrieval from the EHR. These tasks detract from the face-to-face time that’s crucial to the relationship between a doctor and patient.
We’d like to help doctors get back to doing what do best: take care of their patients. We want to help them automate routine tasks and make their workflows more efficient. Our approach is based on introducing ambient intelligence into the clinic, using voice as the primary user interface.
Our first effort is the development of an Intelligent Scribe Service (ISS). We call it a “listening intelligence.” The ISS is a system that listens to doctor-patient conversations, integrates with contextual information from the EHR, and automatically generates a clinical note.
The Intelligent Scribe takes care of the routine work of clinical documentation. It’s adaptive and anticipatory. It provides doctors with context-aware and personalized information. The resulting clinical note is embedded with clinical metadata that unlocks powerful user interactions (such as tracing back the note’s provenance to different parts of the medical encounter). It is a fully semantic note, not just an unstructured text file.
By storing the note’s semantics, we enable simpler information retrieval from the EHR. In addition, we’re able to tap into a virtuous cycle of clinical usage data feeding back into the system, leading to better future recommendations for our system. Additionally, the semantics enable us to trigger intelligent workflows for the physician, the care team, and the patient.
The ISS is designed to be embedded, context-aware, personalized, adaptive and anticipatory – all the characteristics of a successful Ambient Intelligence. Importantly, we follow the principles defined by Microsoft’s approach to AI: a trustworthy AI, with transparency, explainability and accountability at the core, and built with the highest standards for privacy and data security.
In an upcoming series of posts, we’ll discuss further aspects of the system. We’re excited to share our journey with you as we delve into the fascinating world of clinical machine learning and Ambient Intelligence.