Daniel Cane is co-CEO and cofounder of South Florida-based ModMed®, a healthcare IT company that is transforming healthcare through specialty-specific, intelligent platforms to increase practice efficiency and improve patient outcomes.
Founded in February 2010, ModMed has grown to over 1,200 employees and has raised over $332 million in total investment. Known for its progressive growth as a medical technology company, ModMed is frequently recognized both nationally and regionally for its achievements under Daniel’s leadership. In 2020, the company was named one of the Best Workplaces in the Country by Inc. magazine. Between 2016 and 2018, the company was named one of the fastest-growing companies in North America on the Deloitte Technology Fast 500™ list. Starting in 2015, the company has been named annually to the exclusive Inc. 5000 list, a prestigious compilation of the fastest-growing private companies in the country.
Can you share some insights into your background and how it has influenced your work at ModMed?
My journey into tech began during my undergraduate years at Cornell when I co-founded Blackboard. We transformed education by digitizing class notes and creating a platform that gave students and faculty unprecedented flexibility and interaction. For me, Blackboard’s success culminated in 2004 with its IPO, and while our solutions were game-changing in edTech, I couldn’t help but keep an eye out for new challenges.
One such challenge presented itself when I went for a routine checkup with my dermatologist. We had an incredible talk about the struggles of using outdated paper-based systems and ways to fix them. Realizing the bridge between his medical expertise and my technical know-how, we decided to team up and create ModMed along with our first electronic health record (EHR) platform.
At the time, some EHRs already existed, but unfortunately, studies often cited them as one of the leading causes of physician burnout. We took a different approach and designed our EHR to adapt the user experience to the specific workflows of a medical specialty. Our flagship cloud-based EHR, EMA, is and continues to be designed by doctors, for doctors, which has set us apart and defines our secret sauce in the market. Over the years, we’ve expanded our product offerings to include a full suite of solutions that help medical providers simplify and streamline their practice operations and expedite the delivery of care.
How do you see the battle for effective AI in healthcare being won or lost with data?
We’re starting to see a rise in the adoption of AI technology within practices to streamline workflows and maximize efficiency. As we move into an era of using AI to do more sophisticated tasks – such as suggesting treatment or other clinical-support recommendations – it is paramount to have the right data and AI training strategy in place. AI has the opportunity to significantly improve the experience for patients and providers and create systemic change that will truly improve healthcare, but making this a reality will rely on large amounts of high-quality data used to train the models.
Why is data so critical for AI development in the healthcare industry?
Data is the lifeblood of AI, and poor data quality will impair an AI’s performance, leading to suboptimal outcomes. This can have dire consequences in a healthcare setting as patient lives may be at stake. But a more likely scenario is that these negative experiences could undermine both patients’ and providers’ trust in AI, slowing down progress and the positive impact this revolutionary technology can have on healthcare.
For example, in the exam room, AI-enabled ambient listening tools are designed to suggest content for clinical notes for the provider to review and approve. Ideally, this should reduce the amount of time a provider spends documenting within the EHR and allow for more quality time with the patient. However, poor data sourcing and ill-trained AI tools could have the opposite effect, leaving providers to instead spend an inordinate amount of time fixing errors and re-writing notes.
Additionally, bias is a significant risk associated with AI algorithms, and quality data can play a key role in mitigating healthcare disparities. AI models can learn patterns that effectively treat one patient population preferentially compared to other populations, including legally protected groups. By monitoring the data inputs and training on robust and representative data, AI outputs can be more inclusive and accurate.
Can you elaborate on the types of data ModMed uses to train its AI models and how this data is sourced and managed?
At ModMed, we use comprehensive specialty-specific data to help train our AI models with precision. Over the last 14 years, we’ve created specialty-specific, de-identified structured data sets consistent with privacy laws and are now leveraging this in-house data to train our AI models. For example, our ambient listening tool ModMed Scribe has been trained for dermatology, our first specialty launch, on millions of structured parameters from de-identified patient records sampled from a collection of 500 million patient encounters.
How does ModMed define “ethical AI” in the context of healthcare?
The potential for AI to have biases or provide inaccurate information in the form of “hallucinations” or omissions can impact patient lives. For this reason, ethical AI in healthcare is about setting a high standard for accuracy and precision. It means developing algorithms carefully and responsibly and using high-quality and diverse data to help enable more accurate predictions for every user.
Ethical AI is also about ensuring that humans remain in the equation. An AI should not “out doctor the doctor” but instead reduce the administrative burden physicians and their staff experience so they can focus more on helping patients.
What measures are in place at ModMed to allow AI technologies to be developed and deployed ethically?
Our structured data approach—curating high-quality, representative training data sets—helps us make responsible AI a reality. Relevant and de-identified data collected from our EHR systems from a wide variety of practices provides us with a diverse set of training data that reflects different patient populations.
Additionally, our development team embraces data cleaning to facilitate collecting and utilizing high-quality data. This process allows our teams to identify, rectify, and remove inconsistencies, errors, and missing values from the data set. Through this regular maintenance, we can consistently update the AI based on performance data, especially clinical data, where patient outcomes can be impacted.
Can you discuss the importance of transparency and accountability in AI development, especially in healthcare?
Transparency makes accountability possible, which is why it’s such a crucial underpinning to any AI solution in healthcare. Physicians’ top priorities are patient care and safety, so it’s no surprise that 80% of physicians want to know the characteristics and features of the design, development, and deployment of AI tools.
Additionally, not all data is created equal. It’s important to know where and how data is stored and sourced and how regularly it is updated. We’re fortunate that since ModMed’s inception, we have been committed to a data strategy that prioritizes transparency and accuracy. We have a thorough understanding of our data’s sources and quality and are confident that our AI integrations will deliver considerable value to our clients.
How is AI being integrated into ModMed’s specialty-specific EHR systems like EMA and gGastro?
Across our portfolio, we have been utilizing machine learning for some time and strengthening our investment in advanced and generative AI to simplify the business of medicine and expedite quality care. We’re building out an entire AI-powered practice experience that starts before a patient walks in the door, extends through the exam room, all the way through to the billing department.
In the clinical setting, we are in the final stages of our AI ambient listening pilot program for EMA, which we believe will be a game-changer for its downstream functionality and suggested structured content. Our AI-powered documentation solution is designed to streamline the care process beyond just transcription or drafting a SOAP note. Utilizing vast amounts of structured data, we’re training our AI models to capture essential information from doctor-patient conversations and, working alongside our EHR, to suggest relevant content for visit notes, including ICD-10 codes, surgical codes, and prescriptions. This saves physicians precious time and allows them to spend more quality time with their patients.
What specific benefits do specialty-specific AI solutions provide to healthcare providers and patients?
No two medical specialties are alike. They vary widely with the patients they see, the conditions they treat, and the medical codes used for reimbursements. AI solutions must be tailored to accommodate these variations to be effective in any truly meaningful way.
For example, ModMed’s EHRs and AI ambient listening tools are tailored explicitly to each medical specialty, providing highly relevant and precise support to clinicians. Each specialty’s documentation process requires different components within the structured data note, including unique medical codes and terminology. This specialization allows the AI to better understand and anticipate the unique needs and workflows of varying specialty practices, which we believe will result in more efficient implementation, faster adoption, and greater overall effectiveness in improving operational efficiency.
Where do you see the most significant opportunities for AI in healthcare over the next five to ten years?
In the future, AI will undoubtedly permeate nearly every aspect of healthcare in ways we can’t imagine. Already, AI is being harnessed for administrative tasks, and in the near term, this trend will likely surge as AI’s value becomes more apparent.
I also see a future when AI is seamlessly integrated throughout doctor-patient interactions, where the ‘user interface’ or UI is virtually invisible. Instead of today’s screen-based interactions, AI could offer a blend of reality and augmented reality. This future state AI could potentially analyze health records to identify critical insights, predicting a patient’s risk for various diseases. The vast amount of data in medical records presents an opportunity for AI to anticipate future care needs and create and help manage preventive care treatment plans.
This experience could extend beyond the practice setting and become integral to a patient’s daily life. AI-powered wearables could provide personalized support, answer questions, and schedule appointments among other things. AI could also monitor vital signs remotely, detecting and alerting providers to potential health issues. Personalized treatment plans, tailored to individual patients based on data and preferences, could become the norm.
This is truly an exciting time for healthcare. The next five to ten years are ripe with opportunities to further transform the industry and improve the patient experience.
Thank you for the great interview, readers who wish to learn more should visit ModMed.