Rushing Headlong into Healthcare IT’s Digital Past
Rushing Headlong into Healthcare’s Digital Past – A Conversation with Dr. Yin Ho (Part 1)
In the latest episode of Practical AI in Healthcare, hosts Dr. Steven Labkoff and Dr. Leon Rozenblit sit down with Dr. Yin Ho, physician, entrepreneur, and author of Running Headlong: Health IT, Legacy, and the Road to Responsible AI.
This is the first of a two-part conversation that traces the evolution of health information technology over the past 25 years, its transformation (and sometimes distortion) of the practice of medicine, and the lessons we need to carry forward as artificial intelligence becomes embedded in every corner of healthcare. The newest episode of Practical AI in Healthcare features Dr. Yin Ho—physician, entrepreneur, and author of Rushing Headlong: Health IT’s Legacy, and the Road to Responsible AI.
In Part 1 of our two-part conversation, Dr. Ho joins Dr. Steven Labkoff and Dr. Leon Rozenblit to unpack 25 years of digital health transformation—from the dawn of electronic medical records to the market and policy forces that shaped today’s health IT landscape.
Together, we explore how well-intentioned decisions created a fragmented system that prioritizes billing over care—and why understanding that history is essential to building a responsible AI future.
In the latest episode of Practical AI in Healthcare, hosts Dr. Steven Labkoff and Dr. Leon Rozenblit sit down with Dr. Yin Ho—physician, entrepreneur, and author of_ Rushing Headlong: Health IT’s Legacy, and the Road to Responsible AI._
This is the first of a two-part conversation that traces how health information technology evolved over the past 25 years, how it transformed (and sometimes distorted) the practice of medicine, and what lessons we need to carry forward as artificial intelligence becomes embedded in every corner of healthcare.
From the Emergency Department to the Digital Frontier
Dr. Ho began her career as an emergency medicine physician in the late 1990s—just as the internet was beginning to change every other industry. What she noticed, however, was that physicians weren’t part of that transformation. Healthcare seemed insulated from the digital disruption that was reshaping finance, retail, and media.
That realization prompted a career pivot. Dr. Ho left clinical practice to pursue her MBA at Harvard Business School and entered the emerging world of health IT. Over the next two decades, she founded companies, advised startups, and led technology organizations across the pharmaceutical and healthcare ecosystem—including Context Matters, Pfizer, Medidata, Aetion, and Veradigm (formerly Allscripts).
Her book Rushing Headlong is both memoir and analysis—a firsthand chronicle of how the U.S. health IT infrastructure was built, how its incentives became misaligned, and what those design choices mean now that AI is entering the scene.
The “Original Sin” of Health IT
As Dr. Ho explains, the earliest versions of electronic medical records (EMRs) were built for simple record keeping. They were digital filing cabinets—a way to store and retrieve information more efficiently than paper charts. The intent was pure: to make physicians’ lives easier and patient information more accessible.
But as these systems evolved into electronic health records (EHRs), a subtle shift occurred. They became less about clinical care and more about financial throughput. Hospitals and health systems began to rely on digital records for billing, reimbursement, and compliance. Documentation had to support coding and payment workflows, and software vendors optimized for those needs.
“The result,” Dr. Ho observes, “was that clinicians began writing for the billing system, not for each other.” The richness of unstructured narrative data—observations, impressions, clinical nuance—became harder to capture or extract. What was once meant to improve care became a tool to track transactions.
Dr. Labkoff, who built early decision support systems at Brigham and Women’s Hospital, recalls the same turning point. “In the beginning, EMRs were built by physicians for physicians,” he notes. “They were meant to support clinical reasoning. But once payment, compliance, and regulation took over, we lost that focus.”
How Policy Shaped the Marketplace
That misalignment was amplified by policy decisions in the late 2000s. After the 2008 financial crisis, the federal government introduced “Meaningful Use” incentives under the HITECH Act—a well-intentioned effort to modernize healthcare through technology. Hospitals and practices were given five years to adopt certified electronic health record systems or face financial penalties.
The results were swift but uneven. Adoption rates soared, but the market consolidated around a few large vendors—Epic, Cerner, and Allscripts among them. Because certification requirements were tied to government-defined criteria, innovation became compliance-driven rather than user-driven.
“Instead of solving for usability or interoperability,” Dr. Ho explains, “vendors solved for certification. And once the money flowed, the market froze. The same systems we rushed to install are the ones clinicians still struggle with today.”
The outcome, she argues, is an entrenched infrastructure optimized for transactions, not care. Physicians bear the burden of documentation and compliance; patients experience fragmented, impersonal care; and the underlying data—the lifeblood of AI and analytics—is inconsistent, incomplete, and difficult to use for insight.
Two Systems, One Divide
Dr. Ho describes what she calls the “two health IT systems.” One is the clinical record—the data generated at the point of care. The other is the clinical research – not just clinical trial software, but also a secondary data ecosystem—claims, registries, and analytics derived from that clinical record but used for research, quality measurement, or commercial purposes.
The two systems were never designed to work together. Clinical data capture was focused on patient care; secondary data was extracted and repurposed for billing, regulatory, or research needs. Clinical trials data were captured in a completely different data/software ecosystem altogether. The result is a widening gap between the information that clinicians record and the information researchers, policymakers, or AI developers actually need.
Bridging that gap became one of Dr. Ho’s lifelong missions. Through her company Context Matters, she built structured datasets from global health technology assessments—creating a comparative framework for how countries evaluate medical innovations. Her work foreshadowed the rise of real-world data (RWD) and *real-world evidence (RWE)*—fields that attempt to extract population-level insights from clinical care data.
But as she points out, the very systems designed to collect clinical information make this work harder than it should be. Unstructured notes, inconsistent coding, and de-identification barriers all limit how effectively data can be used for learning and improvement.
Lessons for the AI Era
By the end of Part 1, Dr. Ho, Dr. Labkoff, and Dr. Rozenblit reach a shared conclusion: the same design flaws that complicated digital transformation will also complicate the responsible use of AI. Generative AI, large language models, and other data-hungry technologies will encounter the same fragmented data, legacy infrastructure, and regulatory complexity that defined the first 25 years of health IT.
Dr. Labkoff summarizes the transition this way: “We’re standing at another inflection point—like the one we saw when EMRs first appeared. If we don’t learn from that history, AI risks repeating the same mistakes, only faster.”
Coming Up Next – Part 2: The Road to Responsible AI
In Part 2 of this conversation, Dr. Ho will explore how generative AI intersects with the realities of healthcare data—its promise, its pitfalls, and the path toward responsible implementation. From ethical design to data quality and clinical integration, she offers a roadmap for how AI might finally bridge the gap between technology and the human practice of medicine.
Stay tuned for the continuation of this important dialogue on Practical AI in Healthcare.
From the Emergency Department to the Digital Frontier
Dr. Ho began her career as an emergency medicine physician in the late 1990s—just as the internet was beginning to change every other industry. What she noticed, however, was that physicians weren’t part of that transformation. Healthcare seemed insulated from the digital disruption that was reshaping finance, retail, and media.
That realization prompted a career pivot. Dr. Ho left clinical practice to pursue her MBA at Harvard Business School and entered the emerging world of health IT. Over the next two decades, she founded companies, advised startups, and led technology organizations across the pharmaceutical and healthcare ecosystem—including Context Matters, ScienceIO, and Paradigm (formerly Allscripts).
Her book Running Headlong is both memoir and analysis—a firsthand chronicle of how the U.S. health IT infrastructure was built, how its incentives became misaligned, and what those design choices mean now that AI is entering the scene.
The “Original Sin” of Health IT
As Dr. Ho explains, the earliest versions of electronic medical records (EMRs) were built for simple record keeping. They were digital filing cabinets—a means of storing and retrieving information more efficiently than paper charts. The intent was pure: to make physicians’ lives easier and patient information more accessible.
But as these systems evolved into electronic health records (EHRs), a subtle shift occurred. They became less about clinical care and more about financial throughput. Hospitals and health systems began to rely on digital records for billing, reimbursement, and compliance. Documentation had to support coding and payment workflows, and software vendors optimized for those needs.
“The result,” Dr. Ho observes, “was that clinicians began writing for the billing system, not for each other.” The richness of unstructured narrative data—observations, impressions, clinical nuance—became harder to capture or extract. What was once meant to improve care became a tool to track transactions.
Dr. Labkoff, who built early decision support systems at Brigham and Women’s Hospital, recalls the same turning point. “In the beginning, EMRs were built by physicians for physicians,” he notes. “They were meant to support clinical reasoning. But once payment, compliance, and regulation took over, we lost that focus.”
How Policy Shaped the Marketplace
That misalignment was amplified by policy decisions in the late 2000s. After the 2008 financial crisis, the federal government introduced “Meaningful Use” incentives under the HITECH Act—a well-intentioned effort to modernize healthcare through technology. Hospitals and practices were given five years to adopt certified electronic health record systems or face financial penalties.
The results were swift but uneven. Adoption rates soared, but the market consolidated around a few large vendors—Epic, Cerner, and Allscripts among them. Because certification requirements were tied to government-defined criteria, innovation became compliance-driven rather than user-driven.
“Instead of solving for usability or interoperability,” Dr. Ho explains, “vendors solved for certification. And once the money flowed, the market froze. The same systems we rushed to install are the ones clinicians still struggle with today.”
The outcome, she argues, is an entrenched infrastructure optimized for transactions, not care. Physicians bear the burden of documentation and compliance; patients experience fragmented, impersonal care; and the underlying data—the lifeblood of AI and analytics—is inconsistent, incomplete, and difficult to use for insight.
Two Systems, One Divide
Dr. Ho describes what she calls the “two health IT systems.” One is the clinical record—the data generated at the point of care. The other is the secondary data ecosystem—claims, registries, and analytics derived from that clinical record but used for research, quality measurement, or commercial purposes.
The two systems were never designed to work together. Clinical data capture was focused on patient care; secondary data was extracted and repurposed for billing, regulatory, or research needs. The result is a widening gap between the information that clinicians record and the information researchers, policymakers, or AI developers actually need.
Bridging that gap became one of Dr. Ho’s lifelong missions. Through her company Context Matters, she built structured datasets from global health technology assessments—creating a comparative framework for how countries evaluate medical innovations. Her work foreshadowed the rise of real-world data (RWD) and *real-world evidence (RWE)*—fields that attempt to extract population-level insights from clinical care data.
But as she points out, the very systems designed to collect clinical information make this work harder than it should be. Unstructured notes, inconsistent coding, and de-identification barriers all limit how effectively data can be used for learning and improvement.
Lessons for the AI Era
By the end of Part 1, Dr. Ho, Dr. Labkoff, and Dr. Rozenblit reach a shared conclusion: the same design flaws that complicated digital transformation will also complicate the responsible use of AI. Generative AI, large language models, and other data-hungry technologies will encounter the same fragmented data, legacy infrastructure, and regulatory complexity that defined the first 25 years of health IT.
Dr. Labkoff summarizes the transition this way: “We’re standing at another inflection point—like the one we saw when EMRs first appeared. If we don’t learn from that history, AI risks repeating the same mistakes, only faster.”
Coming Up Next – Part 2: The Road to Responsible AI
In Part 2 of this conversation, Dr. Yin Ho will explore how generative AI intersects with the realities of healthcare data—its promise, its pitfalls, and the path toward responsible implementation. From ethical design to data quality and clinical integration, she offers a roadmap for how AI might finally bridge the gap between technology and the human practice of medicine.
Stay tuned for the continuation of this important dialogue on Practical AI in Healthcare.