Founder & CEO of Saykara, a conversational AI assistant for healthcare providers, discusses how it helps reduce the clinical documentation burden while improving the patient experience
1. Tell us how you came to be the founder and CEO of Saykara.
I have been involved with healthcare technology innovations involving speech recognition, machine learning and natural language processing (NLP) for the past two decades. Back in 2000, I co-founded a company, MedRemote, which introduced new speech recognition and machine learning solutions for the medical transcription industry. That company was acquired in 2005 by Nuance, the global leader in speech recognition, and following the acquisition, I spent five years as Chief Technologist of Nuance’s Healthcare R&D division. I chose to exit Nuance in 2011 to co-found another company, Twistle, which developed a highly successful patient engagement and care plan automation platform. I left Twistle in 2015 to found Saykara, but continue to serve on its Board of Directors.
My vision for Saykara is to address a largely unsolved problem in healthcare, which is the clinical documentation burden and data demands put upon physicians in conjunction with electronic health record (EHR) systems. Physicians have been tasked with performing an ever-increasing amount of computer data entry, which stems primarily from insurance billing requirements, public reporting and regulatory mandates. Physicians are not only losing time with patients, that time is being impeded by having to face a computer screen to document the exam, clicking through myriad form fields, tabs, checkboxes and radio buttons and dealing with complex EHR navigation. Numerous studies have shown that physicians spend nearly twice as much time doing administrative work as they spend seeing patients. This results in productivity loss and revenue erosion and is also the number one contributor to physician burnout.
Although current-day speech recognition solutions are adept at capturing a physician’s verbatim dictation, they cannot solve this bigger problem. At Saykara, we have developed a solution that is able to listen to physician-patient encounters, interpret and transform salient content, and create key assets like clinical notes, orders and referrals. This greatly reduces the documentation burden and frees physicians to do what they want to do and need to do, which is to practice medicine and care for patients.
2. Did your doctorate in Computer Science and your stint as a professor at York University set you up for being a serial entrepreneur in the AI health tech space?
What I learned through the process of completing my doctorate as well as my time as a professor is to appreciate the transformative power of computing and the pace at which it changes and evolves. It is critical to always think beyond what is possible today and to contemplate the future and things that have not yet been done. Computing is a field that changes very rapidly, probably more rapidly than any other field, and quite possibly more rapidly than anything else in human existence. If you stop looking ahead at any point, you quickly get left behind. As an entrepreneur, you have to place yourself firmly in a future state and figure out how to deliver it.
3. In what ways is the Saykara platform an advancement over your earlier health tech innovations at MedRemote and Twistle?
MedRemote was born during an era when most physicians were either handwriting their clinical notes or dictating them with a recording device. Although dictating was typically easier and less time-consuming than handwriting, the workflow was more complex and costly in that it required sending the audio tapes to transcriptionists who listened to the recordings and typed paper documents based on what they heard. MedRemote was focused on improving the productivity of transcriptionists by introducing the use of speech recognition technology to that process.
Back then, transcription was considered a $10-$20 billion market in this country, so even a 25% improvement in productivity was considered very significant in terms of potential cost savings. Transcription is, of course, still around today, but is used to a far lesser degree than 15-20 years ago.
Saykara is positioned similarly to MedRemote in that we are introducing new technology designed to improve the clinical documentation process, but the focus now is on interpreting conversations, capturing all of the discrete data that needs to be populated in the electronic health record (EHR), and setting the stage for broader use of artificial intelligence (AI) in clinical processes. Both MedRemote and Saykara can be viewed as cutting-edge innovations of their respective eras. Saykara is not only addressing current-day issues, but we are doing so with current-day tools, technologies and computing power, and we are pushing the boundaries of conversational AI.
Twistle, on the other hand, is not specifically tied to clinical documentation but is nonetheless a complementary technology. While Twistle is focused on improving what happens after the encounter and between encounters, Saykara is focused on improving what happens during the encounter.
4. What are the top three factors that differentiate Saykara from its competitors?
I’d say the foremost differentiator is Saykara’s breadth and depth of experience in this space. I have personally spent the past 20 years developing and bringing to market significant healthcare technology innovations in speech recognition, machine learning and natural language processing, and all members of our company’s executive team are veterans of the healthcare technology industry and possess in-depth understanding and knowledge of physician workflows.
I believe the second most significant differentiator is the approach we are taking with natural language understanding and natural language generation. We have a very sophisticated and flexible platform that allows physicians to use our solution in multiple different modes. Also, the ubiquitous qualities of our platform means it can readily adapt to any area of medical specialty.
The third differentiator is what we are doing in the context of conversational AI. We can understand conversations that naturally occur between physicians and patients, and generate structured data from those conversations in a way no other company does. We are also able to tailor the user experience to each individual physician’s styles and preferences.
5. Can you elaborate on how the NLU (natural language understanding) and NLG (natural language generation) components of the Saykara platform are better than other AI assistants for physicians?
As I described earlier, our platform is designed to understand conversations that naturally occur between physicians and patients. Whereas the speech recognition component of our platform captureswhat is being said, the natural language understanding component interprets what is being said. For instance, if a patient comes in for shoulder pain, the system first has to understand that shoulder pain is a symptom. It then has to build a story around shoulder pain based either on what the patient is saying or the physician is summarizing. When did it happen? What caused it? How severe is it? That is the natural language understanding component, and I think we do that much better than anyone else.
The system next has to take that story in its raw data format and compose a comprehensive, high-quality clinical note, which is what the physician sees. That is the natural language generation component, and I think here, too, we have a significant advantage compared to anyone else, as systems that rely on natural language generation often have problems writing statements that are factually correct or that sound like proper English.
6. Most of your physician users have vouched for Saykara helping them reduce the clinical documentation burden and eliminate after-hours charting. Does your solution also enhance the patient experience?
Yes, our system most definitely enhances the patient experience. With our solution, physicians no longer have to simultaneously converse with patients while looking at a screen and keying data into a computer. We hear time and time again from physicians that giving patients their full attention is critically important, as is eye contact. Otherwise, patients perceive them as disengaged, cold and uncaring, and the patient experience is compromised.
Visiting a physician is when people are at their most vulnerable. To form trusting relationships with patients, physicians need to remain present and be active listeners throughout. In a recent press release, Justin Bayless, the President and CEO of our client, Bayless Integrated Healthcare, was quoted as saying the following about why his provider organization chose our solution – “Whole person wellness defines everything that we do at Bayless. We know that the patient/provider connection is a deeply important part of that equation, and we continuously invest in the tools and technologies to improve that relationship. The more our patients feel heard during their appointments and the more focused our providers are on them rather than data entry, the better our delivery of personalized medicine and the better our patient outcomes.”
We also know that today’s clinical documentation burden eats into visit time. Office visits that used to be scheduled for 20 minutes are now pared down to 10 or 12 minutes, which means physicians have less time to do a thorough exam and patients have less time to talk about their medical conditions, ask questions and comprehend the physician’s diagnosis or care plan. With our solution, we are giving the time and space back to physicians and patients for these things to happen, which leads to better communication, better care and better outcomes.
7. Data is the core of a successful AI tool. How does Saykara sift through the “noise” and process only the most relevant inputs for accurate physician charting?
This is something to which we have dedicated significant attention. Our goal at Saykara is to understand the kinds of information a physician needs to gather based on the reason for a visit, and to predict and anticipate what is relevant and what is not. For example, if a patient arrives at a dermatologist’s office to have skin tags removed and starts talking about the anxiety they have been experiencing over the purchase of a new house, the dermatologist will not likely want or need to document those details. Our system can learn this and focus solely on what is relevant to the dermatologist’s clinical note. On the other hand, if a patient arrives at a primary care provider’s office and starts talking about the anxiety they have been experiencing over the purchase of a new house, the primary care provider will very likely want and need to document those details.
Our system is designed to learn behaviors and patterns based not only on the medical specialty, but also on the individual physician. We currently have a semi-autonomous hybrid model that pairs our AI with a human-in-the-loop reviewer who helps our system make sense of specific types of visits and individual physician preferences, which helps strengthen our models and knowledgebase and ensures we return accurate results to our users from Day 1.
8. Can you give us a sneak-peek into Saykara’s upcoming upgrades or new features? Will you be adding prescriptive/predictive AI elements that aid diagnosis and care planning?
Yes, we will be adding prescriptive and predictive AI elements that aid diagnosis and care planning. Those are central to our mission and we are already starting to add some features along those lines.
By its very nature, our system is “predictive,” as the AI inherently seeks to figure out the context of conversations between physicians and patients. In the not-too-distant future, we will be adding features for ICD-10 and E&M coding as well as nudges to physicians. We will also be looking to queue up diagnostic assessments, order sets and other information that aids physicians in doing their jobs faster, easier and with better results for the patients they serve.
9. What are three ways Saykara has responded to the pandemic and made an impact in these critical COVID times?
The pandemic, as we know, drove a huge uptick in the delivery of remote care, so one of the first things Saykara did in response to COVID was to educate and support our users on optimizing our solution in conjunction with telehealth systems, so they could avoid having to manually document virtual visits. We also developed seamless integration with the Zoom platform and created a COVID-specific model to aid physicians in creating comprehensive documentation for COVID assessments.
Interestingly, the downtime many physicians experienced early in the pandemic provided them the opportunity to research our solution. As a consequence, our sales inquiries went way up, as did the number of new users. Although COVID is still surging, we are seeing patient volumes return to pre-pandemic levels for most of our users and many are experiencing pent-up demand based on months-long delays in elective procedures. This has translated to an urgency around implementing our solution and getting relief from the associated documentation burden.
We are also seeing greater interest from health systems that are seeking not only to fight worsening levels of physician burnout, but to make up for revenue shortfalls by giving physicians a means for seeing more patients. Being able to offload time-intensive and laborious clinical documentation is a win-win on both of those fronts.
11. Besides easy integration with the electronic health record (EHR) and now with Zoom, what other health tech systems can Saykara integrate with to facilitate a larger hospital enterprise management system?
The most important system for Saykara to integrate with is the electronic health record (EHR). All of the data that we create ultimately ends up in the EHR. We also routinely interface with scheduling systems and practice management systems to obtain a physician’s schedule. In some instances, based on the clinical setting, the types of encounters and/or the scope of data involved, a customer may choose for Saykara to interface with something like a lab system or hemodynamic system. Ultimately, we will look to interface with any type of system that contains information directly applicable to the physician or patient experience or otherwise useful to improving the AI and augmenting our knowledgebase.
12. With disruptive health tech now giving rise to robotic surgeries, AI charting, IoT patient care, and more. What further breakthroughs are you expecting in the upcoming decade?
In the future, I envision AI will become central to every aspect of a physician’s workflow. It will very likely be able to identify the best course of care for many of the most common ailments, having learned directly from physicians, and will become the first course of care for many patients. This will free physicians to concentrate on more complex, higher acuity issues. People will look back and wonder how healthcare was ever delivered without these kinds of AI systems, similar to the way people look back and wonder how we ever lived without our smartphones.
13. Which industry other than healthcare are you watching in terms of AI tech innovations?
I am not sure it is a particular industry I am watching as much as the impending revolution of AI systems that can converse in a very natural way with people, which I believe will have an impact across every industry.
When we reach the point that AI systems can understand our conversations – understand the way we naturally converse with each other – this will open up a whole new world of possibilities. It will transcend every aspect of our lives. We are already on the cusp of that revolution, and that is the trend I watch most closely.
14. What is the one leadership motto you live by?
I have never had a motto, per se, but one of the key leadership philosophies I have held over the years is to never confine people to a particular role. I believe in allowing people the freedom to stretch and extend beyond a defined set of responsibilities, then watch and encourage where they naturally gravitate to and where they excel. This is how you get the very best out of people and how they thrive.
Founder and CEO of Saykara
Harjinder Sandhu, PhD, is the founder and CEO of Saykara, a healthcare technology company pushing the boundaries of conversational artificial intelligence and long-form natural language understanding in a mission to reduce the burden of electronic health record documentation and combat physician burnout. Harjinder has been a serial entrepreneur in the healthcare technology space and has stood at the forefront of innovations in speech recognition and machine learning for the past 20 years. Earlier in his career, he co-founded MedRemote, a healthcare-focused speech recognition company acquired in 2005 by Nuance. He then spent five years as VP and Chief Technologist of Nuance’s Healthcare R&D division. He also co-founded the patient engagement company, Twistle, and remains on its Board of Directors. Harjinder holds a doctoral degree (Ph.D.) and master’s degree (M.Sc.) in Computer Science from the University of Toronto and a bachelor’s degree (B.Sc.) in Computer Science from The University of British Columbia. He is a former Assistant Professor of Computer Science at York University and Visiting Scientist at IBM.