Lessons from Dr. Michael Howell, Chief Clinical Officer for Google, on unlocking AI’s potential in healthcare in big tech
Dr. Michael Howell, MD/MPH, Chief Clinical Officer for Google Health, a company dedicated to developing new products and research in close collaboration with clinicians who are experts in their fields
Subscribe to our substack for updates and listen on Apple Podcasts or Spotify. Connect with Andrew or Lipsa if you find this post insightful and want to learn more.
Welcome back to the Pear Healthcare Playbook! Every week, we’ll be getting to know trailblazing healthcare leaders and dive into building a digital health business from 0 to 1.
Today, we're excited to get to know Dr. Michael Howell. He's the Chief Clinical Officer at Google where he leads the team of clinical experts providing guidance for Google Health products, research and services. He has dedicated his career to improving the quality, safety and science of how care is delivered and helping people get the best information across their health journey. He previously served as the Chief Quality Officer at the University of Chicago and was an Associate Professor at Harvard Medical School. He's published more than 100 research articles, editorials, and book chapters and is a nationally recognized expert on patient safety and quality.
In this episode, we talked about why Google got into healthcare, Google’s 3Cs for health, how to create a clinical team within a tech company, the implications of artificial intelligence, and advice for founders in the AI/LLM space.
If you prefer listening, here’s the link to the podcast!
How Google Came into Healthcare
The question frequently arises: "Why has Google entered the healthcare arena?" Every day, Google receives a staggering influx of hundreds of millions of health-related questions from individuals on the internet every single day. In fact, reports indicate that there are more than 70,000 health-related searches conducted on Google each minute. Given the overwhelming demand for accurate and reliable health information, it becomes abundantly clear why Google has ventured into the expansive realm of healthcare. The primary objective is to leverage its vast reach and resources to address this demand and contribute to the pursuit of better health for billions of people worldwide.
Dr. Howell stated that during his grand round presentations he often poses a question to the audience, which typically includes medical students, doctors, and healthcare administrators. “How many of you have gone to the doctor and then gone home to search and double check what the doctor said?” The majority of hands in the room typically go up.
The key to this behavior isn't in the initial verification but rather in the subsequent searches. People return to Google because it proved to be a valuable resource the first time around.
When individuals turn to Google's array of platforms, encompassing YouTube, Search, and other products, a substantial majority—often exceeding 70%—start their healthcare explorations online by seeking supplementary information to address their questions.
“Our goal whether you're seeking health information at any moment or throughout your lifetime, is to give you the best possible answers. This empowers you to comprehend, make informed decisions, and proactively manage your healthcare.”
Furthermore, Google is an engine of innovation. Many foundational concepts in modern artificial intelligence, such as word embeddings and the transformer architecture (the 'T' in GPT), were developed within Google's research facilities. The Google Health team’s primary focus is to explore how these groundbreaking discoveries in artificial intelligence can be harnessed to benefit the field of healthcare. Their mission is to take these innovations and translate them into practical healthcare solutions, contributing to the overarching goal of improving the health and well-being of billions of people around the world.
Google’s 3Cs in Healthcare
Broadly speaking, Google's commitment to making a meaningful impact on people's health revolves around what they refer to as the Three Cs: Consumer, Caregivers, and Community.
First and foremost, they are deeply invested in Consumer Health. Within this realm, they aim to assist individuals in their day-to-day lives. Through platforms such as Search, YouTube, and innovative technologies like Google Lens, they empower people not only to search for health-related information but also to harness sensor data to gain insights into their own health and well-being. This includes the ability to screen themselves for conditions like atrial fibrillation, among others. Their dedication to Consumer Health is about putting health information and tools directly into the hands of individuals, enabling them to take an active role in their health journey.
The second focal point is Caregivers, a term encompassing a broad spectrum of healthcare professionals, from doctors and nurses to health systems and community health workers. Here, they leverage the advancements in artificial intelligence and cloud computing to elevate the quality of healthcare. By making AI-driven discoveries accessible and practical for healthcare providers, they strive to enhance patient care, streamline processes, and improve outcomes. Their goal is to empower caregivers with tools and insights that enable them to deliver the best possible care. This includes even using AI for imaging efforts to help diagnose breast or lung cancer. Furthermore, Google's Care Studio, a centralized platform that enables clinicians to search and gather essential patient information; and this technology now powers Google Cloud’s Vertex AI Search for healthcare.
The third pillar of their mission is Community.
“Where people live, work, learn, and play has a tremendous impact on their health outcomes and well-being.”
To address the health of the community, they focus on various aspects, including the social determinants of health and health equity. They collaborate with public health agencies and researchers to gain a deeper understanding of communities and their unique needs. For instance, they've made anonymized, aggregated datasets available to over 1,000 researchers outside of Google, fostering insights that can drive positive change.
Additionally, their product offerings extend beyond traditional healthcare including health related to one’s environment and surroundings.
“When I woke up this morning and saw on my Google Pixel that the AQR, the air quality index, was low, I remembered that I have a bit of adult-onset asthma and made sure to put my inhaler in my pocket.”
While Google devices can alert individuals to environmental factors that impact their health, the organization also employs data for various health initiatives including reducing auto accidents. Auto accidents are one of the highest causes of death for teens. The focus on community means Google will employ data from sources like Android maps to identify hazardous intersections and reroute users through safer paths, aligning with their belief that health is more than just healthcare.
How to Create a Clinical Team within a Tech Company
Dr. Howell's team operates within Google's Health division, under the leadership of Dr. Karen DeSalvo. Within the broader health-focused ensemble, there are various specialized teams, each with distinct roles. These teams include the clinical unit, the health equity advocates, those dedicated to regulatory compliance and quality assurance, and a team focusing on employee health globally. Among these teams, the clinical unit holds the distinction of being the oldest.
The approach to establishing a clinical team within a tech company has been underpinned by a few crucial principles.
Articulate core values and ensure their alignment throughout the organization. Google has a set of clinical team values you can find here. Values such as making decisions based on rigorous evidence, equity, and ethics are not mere slogans; they are guiding principles. When emphasizing "evidence," it reflects an unwavering commitment to substantiating work with rigorous research and data.
Integrate clinical expertise into the product development process.
“When a product with significant health implications is being developed, we work closely 'at the elbow' with a clinician alongside the product manager and the technical lead. Together, we focus on the myriad decisions involved in creating the product”
Rather than functioning solely as subject matter experts fielding sporadic questions about medical topics, clinicians work side by side with the core product development team. This dynamic involvement allows the clinical team to influence the myriad decisions that collectively shape a product.
Clinical expertise represents the foundation, but it is not the sole requirement. In addition to clinical knowledge, possessing an intrinsic product-oriented mindset, adaptability, and the capacity to seamlessly integrate into a new corporate culture are essential.
It's crucial to acknowledge that the roles within a clinical team at a tech company can vary significantly. The demands of healthcare innovation may necessitate clinicians who not only understand the user and their journey but also demonstrate the ability to be research-oriented, navigate product intricacies, technical challenges, and the dynamic constraints of the business landscape.
Dedicate Resources for Career Advancement of the Clinical Team
Initially, integrating clinical team members into existing roles is necessary, but as the team grows, tailored structures become crucial for retention and career advancement including job ladders and performance review criteria that align with the unique contributions of the clinical team. This investment plays a pivotal role in retaining and nurturing clinical talent, enabling their professional growth within the company. Moreover, such structures enhance the company's ability to attract new team members, particularly those in the early stages of their careers, by demonstrating a clear commitment to offering pathways for increased scope, influence, and responsibility.
Dr. DeSalvo and Dr. Howell published a full article in NEJM Catalyst about this topic and you can read more about it here!
Reducing Misinformation in the Age of AI
As we continue to harness the power of AI and enjoy its numerous advantages, there's a concerning downside - the misuse of AI for spreading misinformation. In today's digital landscape, distinguishing fact from falsehood has become increasingly challenging, especially as generative AI is employed to disseminate misleading information across various domains, including politics, healthcare, and beyond.
At Google, combating misinformation is a constant, round-the-clock endeavor.
“We think about misinformation every single day and every single hour.”
Google does this by employing the 3 Rs.
Remove information that violates their policies
Google places a paramount emphasis on user safety and wellbeing. Content that goes against their policies, especially when it poses an immediate risk to individuals or public health, is swiftly removed. For example, content advocating dangerous or harmful practices like drinking bleach to cure autism is taken down from platforms like YouTube to prevent harm and ensure user safety. This is a proactive step in curbing the spread of misinformation that can lead to real-world harm.
Raise information that is credible and true
In their relentless pursuit of truth and accuracy, Google endeavors to highlight information from authoritative and credible sources by partnering with established organizations in the healthcare industry.
"We have been working for several years with the National Academy of Medicine, the World Health Organization, the Academy of Medical Royal Colleges, and the Council of Medical Specialty Societies. These groups collectively established the principles for identifying credible sources of health information on social media. They emphasize that content produced by national health agencies is more likely to be credible compared to randomly selected internet videos, given the external vetting processes these groups undergo.”
Google has also successfully operationalized these principles across more than 10 countries. Recently, it has expanded its efforts to include individual creators, including licensed physicians, nurses, and mental health professionals in various countries, allowing them to create high quality and verified content for users.
Reduce lower quality content
Google is committed to minimizing the prevalence of lower-quality content through a multifaceted approach that leverages artificial intelligence and strategic partnerships. Collaborations with esteemed organizations like Mass General Brigham, Mayo Clinic, and the American Academy of Pediatrics are centered around producing engaging and accurate content that appeals to viewers. This concerted effort aims to elevate the visibility of higher-quality content, ensuring it takes precedence in search results and recommendations. As a result, users are directed towards information that is not only reliable but also informative and accurate.
The Rapid Advancement in AI and and its implications for Healthcare Providers
The landscape of artificial intelligence is evolving at a remarkable pace, with profound implications across various domains. Within this dynamic landscape, medical question answering has posed an enduring challenge for artificial intelligence. To gauge the strides made in this field, researchers have employed a well-established set of questions, similar to those encountered in rigorous medical licensing examinations, such as the USMLE. While these questions are not exact replicas of the USMLE questions, they exhibit a similar level of complexity.
For years, research teams have diligently worked on improving AI's performance on these questions, gradually advancing at a rate of 2% to 5% per year, inching toward a 50% correctness threshold. Attaining approximately 60% accuracy was widely regarded as akin to passing a medical licensing exam
In December of 2022, Google introduced Med-PaLM, a groundbreaking large language model (LLM) designed to deliver high-quality responses to medical queries. Med-PaLM represents a fusion of advanced AI technology, specifically Google's PaLM 2, fine-tuned to cater to the nuances of the medical domain. The development process involved extensive training conducted by a diverse panel of clinicians spanning the United States, the United Kingdom, and India. These clinicians vetted the model across a variety of parameters including precision, alignment with scientific consensus, low likelihood of medical harm, and the lack of bias.
The first version of Med-PaLM achieved a remarkable milestone by obtaining a 'passing score,' surpassing the 60% threshold on U.S. medical licensing-style questions with an impressive score of 67%. This model not only answered multiple choice and open-ended questions accurately, but also provided rationale and evaluated its own responses.
Fast forward to March 2023, and an updated iteration, Med-PaLM 2, consistently demonstrated an 'expert' level of proficiency in tackling medical examination-style questions. It achieved a remarkable score of 86.5% on USMLE-style questions, marking an extraordinary 18% improvement within a span of just a few months. This remarkable performance far outpaced that of similar AI models and exemplified the rapid strides achieved in the field.
Attaining a high score on USMLE-style questions represents a significant achievement, demanding a blend of medical comprehension, knowledge retrieval, and analytical reasoning. However, what holds even greater importance is the practical application of these accomplishments in clinical practice and healthcare guidance.
To address this critical aspect, the Google research team embraced an innovative approach. They curated a set of commonly asked health-related questions, such as 'Can incontinence be cured?' and 'What foods should I avoid with rosacea?' Subsequently, these questions were presented to Med-PaLM, which, in turn, generated responses to each query. These AI-generated answers were then rigorously evaluated by comparing them against responses crafted by physicians
“In December 2022, physicians who were grading these answers preferred the answers written by other physicians on most dimensions over Med-PaLM. By May 2023, physicians overwhelmingly preferred the answers written by the model on 8/9 dimensions which is remarkable and it is remarkably quick progress.”
The next steps for the Google Health team is to design a multimodal version of Med-PaLM, an advanced system capable of synthesizing and conveying data derived from laboratory values and medical imaging reports, encompassing CT scans, mammograms, MRI scans, and other patient records. This enhancement aims to empower healthcare providers in offering enhanced patient care.
Artificial intelligence is not set to replace doctors; instead, it has the potential to augment their capabilities significantly. Doctors who embrace and adeptly incorporate AI into their practice stand poised to outshine their peers. AI is a tool that will inherently reshape the medical landscape. This prompts essential inquiries into the necessity of integrating AI education into medical training programs, spanning from medical school to residency. Additionally, it underscores the importance of healthcare institutions and practices swiftly adopting AI to bolster doctors' clinical efficacy and patient care.
“In my career ultrasound became a practice that as an ICU doctor you had to learn to be quite good at even though we didn't receive much training beforehand. It is the same way for artificial intelligence where it would be beneficial to learn how to use AI in clinical practice.”
Google’s Partnerships within Healthcare
In recent years, many organizations have forged partnerships with Google Health and Google Cloud. The prevailing trend among large organizations involves a transition from on-premises IT systems and servers to cloud-based solutions. This shift is driven by various factors, including heightened security, cost-efficiency, and improved infrastructure reliability.
Beyond these general advantages, the collaborations between Google and healthcare stakeholders extend beyond traditional data storage and computation, often addressing health-specific challenges through innovative projects. Collaborating with the Mayo Clinic, Google has embarked on an exploration of AI's potential in automating the delineation of healthy tissue from tumors. This initiative aims to enhance treatment planning for patients undergoing head and neck cancer radiation therapy. Furthermore, Google has established a partnership with iCAD, a medical technology and cancer detection firm. In this collaboration, Google is licensing its mammography AI research model, with the objective of integrating it into clinical practices to advance breast cancer detection and the evaluation of short-term individual cancer risk.
These collaborative endeavors are driven by a collective aspiration to improve the health of the individual, the community and the world. The Google Health team firmly believes that artificial intelligence has a pivotal role to play in achieving this goal, and emerging scientific evidence increasingly supports this belief. It's important to clarify that Google Health's AI models are not trained using customer data; instead, they are utilized by customers while maintaining control over their data and training weights. Valuable feedback from users plays a crucial role in refining and optimizing these tools for practical applications.
“Feedback is valuable to us as we contemplate the various ways to transform research tools into production-ready, scalable solutions that benefit people. Our belief is rooted in the idea that by creating exceptional products that significantly enhance people's daily work, or even their businesses, individuals will naturally be inclined to embrace and use our offerings. This approach is in harmony with our core philosophy that people naturally gravitate toward products that demonstrably improve their lives."
In its ongoing efforts to revolutionize healthcare, Google is introducing a new offering through Google Cloud. Customers can now register for early access to the Vertex AI Search features. This advancement empowers healthcare institutions with Google's top-tier generative AI search capabilities, fine-tuned for medical applications. This enhancement enables users to access precise clinical information swiftly and efficiently. Additionally, it facilitates searches across a wide range of clinical data sources, including FHIR data, clinical notes, and electronic health records (EHRs).
Founders Who are in the AI and LLM Space
"Remember that much or most of healthcare is not digital."
Despite the remarkable strides made in AI, Dr. Howell underscores the enduring significance of human interactions in healthcare. For entrepreneurs venturing into this domain, it is imperative to always keep the human element in mind. In the healthcare sector, an established regulatory framework for AI exists, exemplified by the FDA's Software as a Medical Device (SaMD) pathway and the FDA 510K pathway. In areas not covered by these existing regulations, non-clinical products may instill greater confidence in C-suite executives compared to clinical products. When dealing with clinical use cases, founders must consider the enterprise's requirements and preferences, including the choice between having "humans in the loop" or implementing a fully autonomous system.
For entrepreneurs venturing into this space, Dr. Howell offers a pivotal piece of advice: healthcare represents a complex, adaptive system composed of numerous interconnected components. To create impactful AI solutions, it's imperative to consider the comprehensive needs of both patients and clinicians. Neglecting this holistic approach could render AI solutions ineffective in addressing the intricate daily challenges of healthcare.
This goes back to working closely with a clinical stakeholder when developing the product but the crux of the matter lies in the effective utilization of this expertise within the realm of product development. Often, healthcare stakeholders provide input in isolation, detached from product constraints, technical intricacies, and business considerations, but this is less than fruitful. It's akin to having a valuable puzzle piece but being unaware of where it fits into the larger picture.
The key lies in collaborative synergy. A proficient physician, nurse, or psychologist who collaborates closely with a technical team possesses a unique advantage. They comprehend not only the clinical nuances but also the intricacies of product constraints, technical limitations, and business imperatives. In essence, fostering a bi-directional exchange of information among the product team, engineering team, business team, and clinicians yields superior outcomes and enhanced collaboration. Ultimately, this synergy contributes to the development of a superior product tailored to the intricate needs of the healthcare domain.
Advice from Dr. Howell
Clinical training is a great teacher for solving problems
Clinical training is a powerful tool for honing your problem-solving abilities. Whether you've seen a situation a thousand times or encounter a once-in-a-career challenge, clinical training equips you to think critically and find solutions. It nurtures the skills needed to navigate a wide range of scenarios, making you a general-purpose problem solver.
When building a product or delivering a service make sure to focus on the user and make it easy to give feedback
Whether you're in clinical practice, entrepreneurship, or any field, centering your efforts around the needs and well-being of the individuals you serve is paramount. This involves making it easy for your users and patients to relay feedback. By staying attuned to their evolving needs and preferences, you can adapt and refine your offerings to better meet their expectations. This proactive approach ensures that your solutions remain relevant and effective over time.
Embrace diversity of views and be open to new opportunities
Recognizing the inherent value in fostering diversity within your team by embracing the potential it holds to enhance your perspective and problem-solving capabilities. By welcoming colleagues with diverse experiences and expertise, you can fuel innovation and nurture the emergence of creative solutions. Simultaneously, champion a lifelong learning mindset, and be receptive to new opportunities that arise during your career journey. Occasionally, the new and unfamiliar opportunities end up being the most fulfilling
Thanks for reading Pear Healthcare Playbook! Subscribe for free to receive new posts and support our work. If you found this post insightful or want to learn more, please feel free to connect with Andrew or Lipsa.
Interested in Google Health? Learn more on their website and LinkedIn
A note from our sponsor: PacWest
Looking for guidance, connections, resources, opportunity? Pacific Western Bank’s banking products and services are built to support your evolving needs as you navigate the challenges of growing a successful business. As you continue to scale, our team will be with you every step of the way. Ready to take your business to the next level? Learn more: pacwest.com
 | A guest post by
|