In the space of one lifetime, medical information technology has transformed radically. Penn’s chief medical information officer is living that lifetime.
By S.I. Rosenbaum
Photos by Peggy Peterson and Graham Perry
A fountain pen and an index card were two of the most essential assets in medical information technology fifty years ago.
Clarence William Hanson, Jr., MD’55, the director of the Emergency Ward at the Hospital of the University of Pennsylvania, would sit with patients, carefully taking notes on the index card before slipping it into his jacket pocket and proceeding to the exam. Sometimes, his son, Bill played with the fluoroscope nearby. Later, in the evenings, he’d settle in at home with the baseball game on the radio, leafing his way through the index cards and narrating encounter summaries into a dictaphone.
Those tapes then went to his secretary, who plugged them into a transcription machine with headphones and foot pedal. The transcribed patient records then went into a Pendaflex folder and then into a file cabinet.
That was the state of medical information technology when young Bill Hanson was growing up: paper and pen, onion-skin paper.
Today, as Penn’s chief medical information officer, C. William Hanson III, MD’83, is on the other side of an information revolution as both a physician and onetime computer coder. He works in a glass office in the Perelman Center for Advanced Medicine, at the heart of a vast network of information—a hive of interconnected computers and smart devices adding up to a system that spans Philadelphia and far beyond.
Medicine has always been about information: gathering it, interpreting it, connecting it to other information. What has changed so far in Hanson’s lifetime is not just the tools—the fountain pen and filing cabinet haven’t been simply replaced by computer keyboard and database. Medicine, Hanson says, “is in the process of transitioning from a scientifically informed art to what hopefully will be an artistically informed science.”
Hanson is a member of the last generation of people who were self-taught computer programmers and tinkerers. He has witnessed every stage of this rapid evolution—or occasionally just found himself unwittingly adjacent to it, like a medical information Forrest Gump.
Sneakernet and Silicon Valley
One of the first places in Penn’s medical enterprise where computer technology was implemented was in financial records.
That’s where Hanson wrote some of his first code. He’d learned how to program in high school, so after college and through medical school he wound up working as an “odd jobs guy” under HUP’s director of data processing, Rich Viale, PhD. One of Hanson’s odd jobs was to write a program that would check the hospital’s ledger of checks written against the bank’s ledger of checks cashed.
Both ledgers were digitized, meaning that, at the time, they were stored on magnetic tape. There was only one problem: the only way to get the punch tape from the bank to the hospital and back was on foot. So Hanson found himself sprinting the five blocks regularly, creating what old programmers would call a “sneakernet.”
“I was a kid running this business-critical mission, this critical operation, lugging these tapes around in a knapsack between the bank and the hospital,” he says.
But the era of paper punch tape, punched cards and magnetic tape was almost over, and there were those who were already looking ahead to what might be possible with a new generation of lighter, leaner, more networked machines. As a medical student in 1980, Hanson helped host a talk by one of those visionaries: a former Penn biochemistry student named Larry Weed, who was then teaching medicine at the University of Vermont.
Weed had noticed that hospital records were often illegible and disorganized—as one writer put it, “highly variable in comprehensibility”—and in 1968 had published a paper in the New England Journal of Medicine that proposed a new way of consolidating the scattered information. “He was essentially talking about a structured way of recording medical encounters that would be susceptible to computerization,” Hanson says.
Weed’s problem-oriented medical record, or POMR, was a way to keep all the information on a patient together, from one visit to another, so different physicians could check each other’s work. He first described what is now known as SOAP (subjective, objective, assessment, plan) note charting. And he envisioned the system as being linked to a database of medical knowledge that could help guide a doctor’s decisions.
Meanwhile, across the country in Palo Alto, a company called Apple was working on its version of the invention that would eventually implement Weed’s vision: the personal computer.
In 1983, Hanson arrived there for his first residency (in Internal Medicine) at Stanford University Hospital, but he had no idea how close he was to history.
“The computer revolution was happening blocks away from me that I had no idea of,” he says. “I was oblivious to what was happening with Apple. I was frankly oblivious to what was happening academically at Stanford.”
Ironically, his clinic attending physician was Edward Shortliffe, MD, PhD, who would go on to become one of the pioneers of early medical informatics and artificial intelligence. In 1975 Shortliffe had designed a rule-based computer program, MYCIN, that could assist doctors in diagnosing infectious disease.
MYCIN was never put into practice, out of concerns about liability and cost. Still, it was far ahead of its time, and remained a foundational experiment in medical artificial intelligence (AI).
Hanson, the frazzled young resident, knew none of this. To him, Shortliffe was just his clinical attending. “I was sort of aware of the fact that he and his colleagues had developed some software,” he says. But his future career would bring him back into Shortliffe’s orbit.
Shortliffe, for his part, remembers Hanson as a medical student “on the cutting edge in terms of demonstrating that not everybody needed to look like a grey-flannel-suit guy.”
As an attending, Shortliffe knew that Hanson and his peers were absorbing information “firehose style.” But he always tried to drop in some information about what computers and medical informatics could do, just to make sure they were familiar with the concept.
“He left knowing the word, which is a lot more than most doctors in those days,” Shortliffe says.
Tech Hub in the ICU
Even in 1986, operating rooms at HUP were full of technology—and it was smart technology, Hanson realized, as he began his second residency in anesthesia. Oxygen and anesthesia might flow into a patient, but data flowed back out. “All of these [devices] were, to a degree, computer enabled,” he says. “They had data that they were generating, and that you could access with a computer.”
He wondered what kinds of questions that data might be able to answer.
He started out with a cardiac monitor. By “hacking” the device, extracting and analyzing its data, he was able to write a program that could use that data to predict a patient’s deterioration earlier than a bedside nurse. He was also able to disprove a widespread belief that giving heavy smokers extra oxygen would slow down their breathing, causing CO2 levels to rise. By checking the data, he found that breathing rates remained the same, and another factor was causing the CO2 increase.
“I was able to debunk prevailing wisdom and then bring that to a computer model and help explain what was really going on,” he recalls. “The significance of that is totally current—the idea that you could take something and say, ‘Let’s model this problem with known science and try different inputs and outputs and come up with new explanations and new therapies.’”
HUP wasn’t the only place where doctors were becoming interested in the medical potential of technology. In 1999, two intensive-care doctors at Johns Hopkins Hospital approached Hanson and asked if he wanted to start a company with them. They were looking to apply the emerging field of telemedicine to the ICU, so that a remote intensivist could monitor patients 24 hours a day.
“I was tempted,” Hanson recalls. But he took a counter offer to remain at Penn instead, which afforded him a chance to take a sabbatical in the computer science department at Princeton. “This was the most spectacular opportunity to be a student again, in a place where there was so much intellectual richness,” he says. He continued teaching a course on medical informatics as a visiting professor there even after he returned full-time to Penn.
Some time later, the chief medical officer called him in. “Would you like to be director of a tele-ICU?” he asked.
Penn, it turned out, had decided to contract with a company that would eventually be known as Visicu—the very company Hanson had declined to join a few years earlier. “Poetic justice,” he says. “It’s like we had a bad date and then we end up getting married.”
So Hanson set out to help Penn build one of the very first telemedicine-enabled ICUs.
It was fitting for this happen at Penn, which had already been a pioneer in the field of telemedicine. In the 1940s, Jacob Gershon-Cohen, MD’24, an assistant professor of Radiology at Penn’s Graduate School of Medicine, noticed that newspapers had obtained the ability to transmit photographs across telephone wires using an early version of the fax machine.
Gershon-Cohen found he could use similar technology to send radiological images via telephone from Chester County Hospital back to Philadelphia for analysis by expert radiologists. He called the technique “telegnosis”, and it saved lives: for example, in one case doctors at Einstein Medical Center were able to diagnose intestinal obstruction in a patient at Chester County via an emergency tele-consultation.
He even experimented with “videognosis,” transmitting radiographs using a television camera and a shortwave radio.
So telemedicine was not exactly a new idea—but no one had yet smoothed out the wrinkles in using it to create a constantly-surveilled ICU. At Penn, Visicu’s system set up a remote station where an intensivist could see data from each patient’s bedside monitor, as well as the view from a camera mounted on the wall in the patient’s room. The cameras were moveable and so high-resolution that the doctor in the remote station could tell how dilated a patient’s eyes were. The station was also equipped with automatic alerts if any patient’s vitals dipped, as well as software that would walk doctors through possible responses.
There was a sharp learning curve for all involved. “It was very challenging bringing Big Brother to look over the shoulders of the people actually in the ICU,” Hanson recalled. As for him, “I didn’t really know how to practice medicine through a camera.”
Initially, the remote station covered two ICUs: the one in HUP, which Hanson ran, and the one at Pennsylvania Hospital. The staff at HUP knew Hanson, and they were already at that time “innovation oriented,” he says. At Pennsylvania Hospital, the transition was rougher. It wasn’t unusual, he remembers, to find lab coats hung over the cameras.
One night, a patient had to be intubated. As the procedure began, Hanson—watching through the camera—saw that the nurse anesthetist was having some trouble. If he’d been in the room, he would have automatically moved to help her, but he wasn’t in the room. Still, he thought, maybe he could make suggestions, the way he would if he were overseeing a resident.
Visicu had equipped the patients’ rooms with not just a camera, but also a speaker that would allow the remote doctor to talk directly with nurses. So Hanson turned on the microphone and made his suggestions.
He saw the nurse freeze, then look slowly around the room.
“It became apparent to me that she had no idea who I was, or where this voice was coming from,” he says. “This is a perfect metaphor for this new way of doing medical care.”
Slowly, though, it became clear that the extra eyes of the remote intensivist were keeping fragile patients safer. The tele-ICU grew from covering 35 beds, to 70. Today, 270 beds at Penn are monitored by a remote intensivist—as part of Penn Connected Care, one of the largest telemedicine programs in the country—and nationwide, Hanson says, roughly 20 percent of ICU beds are monitored remotely.
“It’s one of those things that’s initially met with skepticism and it’s grown into being adopted nationally,” he says—one of computerized medicine’s “first real successes.”
Captain of the Sea Change
If telemedicine was disruptive, it had nothing on the electronic health record, or EHR.
The EHR has become central to how doctors practice medicine today. In addition to digitizing a hospital’s paper files, they also standardize doctors’ case notes, the way Larry Weed had foreseen, and incorporate AI-assisted decision making, the way Ted Shortliffe’s MYCIN did. A well-designed EHR can prevent physician error, save money and improve outcomes; a poorly designed one can burn out doctors and kill patients.
The sea-change began in 2010. That year, the federal government established a “meaningful use” incentive program: to be eligible, a hospital had to create an EHR that met government standards by 2014. At Penn, Hanson was tasked to lead the effort as chief medical information officer.
Today, as CMIO, Hanson oversees all the data exchanged with other hospitals and insurance companies, and works to find ways to use that data to create better care; he oversees Connected Care, from the tele-ICU to tele-behavioral health; and he oversees the development of mobile apps to help patients and doctors keep track of treatment.
But building Penn’s EHR—what would eventually become PennChart—has been “job one,” he says.
“I won’t say it was a baptism by fire,” he says. “But we had no predicate.”
Across the country, hospitals and medical practices struggled to meet the government’s EHR standards. The implementation of EHRs was necessary—medicine was far behind other industries in converting to digital records—but it was also expensive and onerous.
“It's taking a very, very complicated industry and replumbing it—while you're still operating it,” Hanson says. “You can’t just shut down the highway; you have to keep taking care of patients.”
Not every hospital was able to recoup the cost quickly enough to keep from folding, Hanson says. “Some places were brought to their knees by the implementation of electronic records,” he says. “Some places had nine-figure hits to their bottom line.”
In 2005, Penn sociologist Ross Koppel, PhD, published a paper pointing out that bad EHRs could cause more problems than they solved. “I said, not that the emperor had no clothes, but that he was pretty friggin’ threadbare,” Koppel says.
Koppel studied Penn’s system for inputting prescriptions. He identified 22 possible sources of user error that could be potentially deadly, writing: “As hospitals and clinicians implement these systems, they must consider the errors [EHRs] may cause, as well as the errors it may prevent.”
Even an EHR that didn’t create errors could simply annoy doctors—in some cases, causing them to retire rather than face the computer every day. “It’s called ‘death by a thousand clicks,’ Hanson says. Doctors complained that they spent more time satisfying the demands of the EHR than with their patients. “The sound of medicine is not the click of a mouse. It is the human voice,” lamented one group of doctors in an op-ed.
Hanson’s role as CMIO, he soon found, was similar to a factory foreman—a go-between. “They needed to have someone who on the one hand was interpreting what the IT community was saying to doctors, and on the other hand what doctors were saying to the IT community,” he says.
It took four years just to get the EHR up and running; Hanson likens the process to “a snake ingesting a large meal.” But now, Penn is concentrating on improving it. Last year, Penn launched an initiative to substantially redesign and enhance the EHR, with the goal of making it more intuitive and user-friendly, like Amazon, Facebook or Netflix. Hanson’s team is working with information services, the EHR transformation team, and the Penn Medicine Center for Health Care Innovation, including the Nudge Unit and the Center for Digital Health, to achieve this goal.
Hanson is looking further ahead. The EHR isn’t just a useful clinical tool: like the operating-room devices he hacked years ago, it can yield up a treasury of big data. And with big data comes the possibility of effective AI.
“The number of things you can do with digitized data is proliferating every day, and it’s really going to impact medicine writ broad in the coming decades,” he says.
An experienced doctor might see hundreds or even thousands of patients with a particular condition over the course of their career. “But an AI can say, metaphorically speaking, ‘I have seen a million patients,’” Hanson says. The pattern recognition powers of AI “may contribute to or disrupt entire medical specialties,” he adds.
AI can also profile patients to help doctors anticipate their needs—in the same way Amazon analyzes someone’s purchases and suggests new books, data from the EHR can allow AI to analyze a patient’s history and suggest services they might need in the future. The predictive health care team at Penn Medicine under Hanson’s purview for the last several years has already developed algorithms that aid in decreasing readmissions for at-risk heart failure patients and in proactively offering palliative care consultations for patients who are most critically ill.
Hanson is looking forward to seeing where else the technology takes medicine. Wherever that is, it will no doubt be as different from today as his father’s pen and notecard.
“I would say the pace of change is intensifying,” Hanson says. “We're nowhere near the end of it.”