Martha Dominguez de Gouveia / Unsplash / Undark

It was a cool April day in 2016 when Michael Champion’s wife, Leah, noticed that her husband’s forehead was drenched in sweat. She took his temperature and couldn’t believe the number: 102.4 degrees Fahrenheit. He had been lying in bed more often, and as she watched him over the previous weekend, she’d hoped he would improve. But when Michael became so weak he couldn’t even hold up his head, she knew this couldn’t wait any longer, and she called an ambulance.

I was a first-year resident working at the Palo Alto veterans’ hospital when I got a call to evaluate an elderly vet with fevers. Michael was not able to tell me his symptoms. From his chart, I gathered that he was a long-standing diabetic. Several years earlier, he’d also suffered a stroke that left him with a weak right side, slurred speech, and trouble swallowing. He had a permanent feeding tube and required bladder-catheter insertions several times every day. Stripped of the ability to interact the way he could before his stroke, he spent most of his time in bed or in his wheelchair. Leah was his primary caregiver.

Now she sat at his bedside, coloring in the details of a medical history he was unable to voice himself. She told me how tired and disoriented he seemed. He was pale. It wasn’t right.

I ordered blood and urine tests that traced his fevers to a multi-drug-resistant infection in his bladder. We treated the infection with antibiotics and worked on techniques for hygienic catheterization. Because of the infection, his blood sugar ran consistently high, so we also added extra insulin to his diabetes regimen.

He was doing better within a few days, and his mental status had perked up. Every morning, when I asked how he was feeling, he was able to provide one- or two-word answers. Several times he gave me a thumbs-up.

But his hospitalization had taken a toll—especially on Leah, who now realized Michael would need to regain his strength for her to care for him at home. He could hardly move from his bed to a chair without two people assisting him, and even that left him drained. Our case manager identified a skilled nursing facility nearby, just south of San Francisco, with continued physical therapy and around-the-clock nursing care. I remember Leah expressing relief about the choice. She would be able to visit him every day but still rely on a dedicated team of professionals to help him until he fully recovered.

I performed the rituals of hospital discharge that had become second nature to me as a resident. I typed up a discharge summary outlining each of his medical problems. I spelled out his antibiotics plan. I wrote out his new insulin regimen—an additional injection every six hours and extra doses with his tube feeds, on top of his usual morning dose. I summarized what we were thinking, what we had done, and what needed to be done next. Whenever possible, I had learned to bolster that sheet. I used simple and straightforward language. I bolded. I double-checked my medication list. I knew this sheet was often the only guidance a nursing facility would receive. If I didn’t write something here, it very often didn’t exist.

But I also knew that even if I did write it down, it might not be read by the caregivers and health-care professionals who would treat him next. And I knew that only fragments of their notes and charts were likely to get passed down the line of Michael’s care, too. That’s precisely what happened. Over the next few weeks, Michael would return to my hospital more than once, in bad shape as a result of unconnected records that were not easy to transfer. In one instance, he didn’t receive the insulin doses that I had so carefully marked in his discharge form, which left him in a near-comatose state.

Michael isn’t alone. Every year, an untold number of patients undergo duplicate procedures—or fail to get them in the first place—because key pieces of their medical history go missing. Countless others suffer from medication errors. Hospitals, nursing homes, and other medical facilities use a patchwork of methods to track records, relying on proprietary technology or old-fashioned communications such as faxes and paper notes. These systems don’t always sync, and the collective costs to patients, hospitals, and the economy as a whole are impossible to quantify—although some experts say consistent and cohesive health-information technology could save billions of dollars. An initiative from the U.S. Department of Health and Human Services aims to unify these disparate systems, but we remain far from a universal electronic medical record that would solve the problem.

Meanwhile, we have stories like this one, which is a story about gaps. Michael would slip through these gaps. But as I filled out his record for the first time that day in 2016, neither Leah nor I knew just how short our best efforts would fall. Back at the hospital, she and I shook hands, and she thanked me for my care.

More than a year and a half later, I asked Leah how she had felt at this point. She said, “I had my concerns about him leaving.” But we had faith in the system and she was relieved that her husband would be in a skilled nursing facility, receiving help for medical problems that had become too overwhelming for her to manage alone.

He would get stronger soon, she thought. Then they’d be able to go home.


The U.S. health-care system is dynamic by design. Patients move from one hospital to another; from a hospital to a rehab facility; from the wards to the intensive-care unit; from the hospital to a primary-care setting. These transfers are inevitable, as a person’s health either improves or declines, or if that person simply desires a second opinion. In nonemergency situations, it’s somewhat of a medical free market. Patients have every right to take advantage of it. But when the patients move, their histories often lag behind.

When I meet a new patient, I have to gather slips of these histories from various sources—electronic records, paper documentation, outside faxes, notes in wallets, family members—to piece together a meaningful narrative. Why is this person on a steroid medication, and is there a plan for tapering? Is this poor kidney function new? How had Michael’s fluctuating blood sugar been managed when he’d been sick before?

While most hospitals in the United States today use electronic health records, they remain disparate, with hundreds of different interfaces and minimal data sharing from one care facility to the next. Today, less than half of hospitals electronically integrate data from other hospitals outside their system, and only 30 percent of skilled nursing facilities share data outside their walls, according to the Office of the National Coordinator for Health Information Technology (ONC), a division of the U.S. Department of Health and Human Services.

Without an easy way to get a patient’s full medical files, I must ask where their previous doctors were, have the patients sign a release form, fax it to the other hospitals, and receive stacks of papers in return. Then I dig in. Unable to use control-F on a stack of paper, I sift through several to hundreds of pages to find the few values of importance. If I’d like to see images, such as MRI or CT scans, it’s more involved: I request CDs, wait for them to be mailed over, walk them to our radiology department, fill out another form, and then wait another day or so until I can see them in the computer system. The more places the patients have gone, the more there is to unravel.

When a patient with a complex medical history like Michael’s arrives under my care, it’s like opening a book to page 200 and being asked to write page 201. That can be challenging enough. But on top of that, maybe the middle is mysteriously ripped out, pages 75 to 95 are shuffled, and several chapters don’t even seem to be part of the same story.

Meanwhile, everyone around me is urging: Write now.

Mere days after the Champions left, my senior resident asked me to evaluate a new patient. He had a history of diabetes and stroke, and he was so tired he couldn’t keep his eyes open. “Admit to medicine,” the note from the emergency room had read. When I pulled back the curtain, my heart sank. The patient was Michael Champion. His frail body lay still in the hospital bed, eyes closed, unable to communicate. When the emergency doctor gave Michael’s sternum a brisk rub, he awoke only to instantly fall back asleep.

Leah was next to him, her face contorted. She told me that things had gone wrong the moment they had arrived at the nursing facility. The scheduling of his tube feeds was disrupted by the antibiotics regimen. The extra insulin I had written in his chart had also been withheld.

A blood test revealed one likely source of Michael’s lethargy: His blood sugar was nearly four times as high as normal. A reading just slightly higher can tip a diabetic into a condition called hyperosmolar hyperglycemic state, with risks of extreme dehydration, electrolyte imbalances, and in the worst cases, brain swelling. Though brain swelling is rare, its mortality rate is high. I spoke to Leah at the bedside. Incredibly, she wasn’t angry. She knew the problem was fixable, and she relied on us to fix it, just as we had his infection. But she was confused. The entire time Michael was in the hospital, he was given extra doses of insulin every morning, afternoon, and evening. At the nursing facility, he had been given none. How could this have happened?

Leah’s question gets at the million-dollar one: Why?

The answer lies in the tangled evolution of e-health technology. In 2004, President George W. Bush created the ONC within the U.S. Department of Health and Human Services. In 2009, Congress then authorized and funded related legislation known as the Health Information Technology for Economic and Clinical Health Act (HITECH) to stimulate the conversion of paper medical records into electronic charts. And indeed, many hospitals and doctors’ offices did this successfully, says Karen DeSalvo, the national coordinator of health information technology from 2014 to 2016, by “digitizing the care experience of every American.” But electronic-health vendors made proprietary systems that weren’t always compatible with one another, which made it hard for records to transfer between medical facilities. Now, says DeSalvo, “we need to get them blended.”

In 2014, DeSalvo’s office had about $100 million to hit this goal. But even with this level of support, “that kind of transformation is difficult,” she says. There are hundreds of different vendors using diverse technological platforms, each with a unique way to organize patient data. Finding a universal way to blend the systems has been a bureaucratic and engineering nightmare. Sometimes the barriers are intentional, DeSalvo says, as vendors don’t necessarily want to make it easy to share data with hospitals using competitors’ systems and may charge them a fee to do so. Blocking also occurs because of a misunderstanding of patient privacy rights. As efforts to exchange data go public, some critics worry about security leaks.

The hand-wringing over privacy may not be necessary. According to Lucia Savage, a lawyer and the former chief privacy officer at the ONC, the same privacy rules apply no matter how patient files are shared—whether by handwritten notes, faxes, or electronic files. “Doctors, nurses, physician assistants—they have ethical rules,” she says. “They’re not supposed to be snooping around someone’s data because he’s Steve Jobs ... We have to assume people act professionally and ethically in this space.”

For patients, the trust often already exists. When anyone comes into any health-care facility, we review all the records we have. Together, we openly discuss and dissect deeply personal information. In fact, patients may be the biggest advocates for sharing medical information, says Mark Savage, the director of health policy at the University of California at San Francisco’s Center for Digital Health Innovation. (He’s also married to Lucia Savage.) Patients tend to take it for granted that their doctors are talking to one another, he adds, and “they’re frustrated when information is not exchanged.”

There are data backing up these ideas. A 2014 survey of more than 2,000 patients by Savage, then with the National Partnership for Women and Families, and colleagues showed that 95 percent felt that electronic records “were useful in assuring timely access to relevant information by all of their health-care providers.” And more than three-quarters said that they already share information with their health-care providers all or most of the time. Those who used electronic records compared with paper records reported more trust in their providers to protect privacy.

These results resonate for me. Over my last few years as a doctor, I can’t think of a time when a patient complained that a doctor knew too much of his or her medical history. Yet many times, I’ve heard frustration that we don’t know enough.


In the absence of a system that reliably transmits information, we get creative. When the system fails to fill in the cracks, we hope the patients can help. I ask them: What did the doctors say? What did the testing show? Was it a loud machine where you lay flat, or did someone use a probe coated with cold gel?

When patients don’t know, the burden of creativity falls on motivated bystanders—sometimes other health-care providers or family—who, through grit, persistence, and clever work-arounds, find ways to cobble together information and pass it forward.

Michael Champion was fortunate to have a living, breathing medical record in Leah. Early on, when we met, I noticed how she would pull a yellow legal pad from her purse. Later, I learned that the pad accompanied her wherever Michael went, and she used it to jot down the minutiae of his care. She spoke with a deep knowledge of his medical issues, and hospital staff would often ask her if she was a doctor or nurse. No, she would say. This was a role she had acquired out of necessity.

By the time we lowered Michael’s blood sugar and he was ready to be discharged for the second time, I learned my lesson. I had performed my rituals of discharge diligently. But I needed more safeguards. When in doubt, duplicate. I had a plan, and it had three parts. One: I would print my discharge instructions, as before. Two: I would call the nursing facility and verbally pass off Michael’s care plan. Three: I would rely on Leah. As the only source of continuity in Michael’s many transitions, Leah would be the glue. If everything else fell through, she could advocate.

Michael Champion in June 2016 (Courtesy of the Champion family)

Many patients don’t have a Leah, though I’d seen a lucky handful who had people go above and beyond to fill in the gaps. There was the oncologist who needed the biopsy results for one of her new cancer patients. The biopsies had been done long ago, at another hospital, and the patient couldn’t remember who had done them or what they showed. So the oncologist cold-called all the physicians in the department; when none of them knew the patient, she asked for the names of all the recently retired doctors and called them, too, until she got what she needed.

I also remember the Italian man whose clever physician spared him an invasive procedure into his heart from 6,000 miles away. The monitors capturing the man’s heart rhythm showed ST elevations—the scary sign of a serious heart attack colloquially known as the “widow-maker.” The patient saw the concerned looks on our faces as we watched his monitor and then calmly reached into his pocket and pulled out a folded paper that had been cut to wallet size. It was an old rhythm strip, labeled 10 years ago, and it had the same ST elevations. His primary-care physician had told him to keep it on him at all times, so that any new doctor would learn he had an irregular baseline and not wheel him straight to the cardiology lab whenever he walked through a hospital door.

I thought of these stories on the morning of Michael’s discharge. He was headed to a different nursing facility, about 10 miles south of the first one. I spoke to Leah candidly. I told her disappointing truths: Discharges are the most dangerous times of a hospital stay and I had seen plans break down even at the best institutions. Because of that, I said, I am going to tell you the details of his medical care. The nursing facility should already know these things. But if things go wrong, you can fill in the gaps.

It was unfair. Beyond dealing with the emotional weight of Michael’s illness, I was asking her to carry his nuanced medical care as well. I was asking her to perform the jobs that the medical system around her was supposed to do. But no one was incentivized to care as much as Leah and she took on the challenge with enthusiasm.

We spent the next 30 minutes talking shop. She asked questions. I answered them. She took notes on her yellow sheets. We brainstormed, together.

“I want to empower you to advocate for him,” I said.

She put down her legal pad, reached out to shake my hand, and then, reconsidering for a moment, instead softly wrapped her arms around me for a long hug. She was about to head into the medical unknown.


The next time I saw Leah, she sat by her husband’s side as he lay in a hospital bed again. It was one week later. We were in the ICU. She rose from her seat when she saw me. With every visit, I swore I could see new wrinkle lines. It was as though she was in a time warp, aging a year through every week of her husband’s medical decline.

“It was like you were prescient,” I recall her saying. “I did everything you said, and it still didn’t work.”

This time, the gap was in Michael’s tube feeds. Through his feeding tube, he was meant to receive both nutritional formula and water. Water is critical to balancing electrolytes in the bloodstream; without it, the relative sodium levels in the blood can rise. The cascade of side effects increases as the sodium does: confusion, seizures, coma, death. But the second nursing facility hadn’t given Michael the right amount of water. His sodium soared and, once again, he was nearly comatose.

Leah had alerted the physician. This is completely different, she pointed out. He’s not speaking. To which she said she was told: This is just the progression of his illness. He’s a sick person, you know.

But Leah was right—though she didn’t know just how sick Michael was. She came back from running errands shortly thereafter and found Michael being loaded into an ambulance. In the emergency room, his labs showed a sodium level nearly 20 points higher than the upper limit of normal. It was infuriating. Michael was now in the ICU, less than three weeks after he was admitted and discharged and admitted and discharged again. It was preventable.

Leah’s question was potent: “How could this have happened?” At the time, I imagined many possible answers. Maybe my printed discharge summary had gotten lost. Maybe his initial nurse left it out during a verbal handoff—where the responsibility for care is transferred from one provider to another—to be forever erased from his history.

More likely, a nurse or doctor had held a handful of pages in their left hand and a handful of pages in their right, trying to re-create a medical story from scratch. And the details of his sodium management were lost in the shuffle.

In 1999, the Institute of Medicine, now called the National Academy of Medicine, published one of the most famous reports in medical history. “To Err is Human” noted that between 44,000 and 98,000 people die in hospitals each year from preventable medical errors—a figure that was later compared to the equivalent of a jumbo jet crashing every day. More recent studies suggested this number was too low, raising it to more than 400,000 deaths per year. If they were a disease, medical errors would now rank as the third leading cause of death in the United States, a 2016 analysis published in The BMJ found, right behind heart disease and cancer.

I wanted to find data on errors that occur as patients move across systems. What is the cost of these transitions? I imagined both errors of commission (of repeat lab tests, scans, and even procedures) along with errors of omission (delayed or missed diagnoses because of an incomplete medical context).

The data we do have is unsettling to read. The Joint Commission, a large accreditor of health-care organizations with a focus on patient safety, studied transitions of care and concluded that they are largely ineffective, leading to adverse events, hospital readmissions, and soaring costs. Their conclusions were supported by several key studies. One reported that nearly 20 percent of patients experienced adverse events—like Michael Champion’s sugar and sodium spikes—within three weeks of discharge. Almost half were deemed preventable. And when patients are admitted to hospitals, more than half have at least one discrepancy in their medications.

How much is faulty record sharing to blame? My colleague, Marta Almli, an internal-medicine doctor, surveyed the resident physicians at my hospital and others and found widespread dissatisfaction with how we obtain medical records: Among 58 physicians surveyed, 81 percent said it was “somewhat difficult” or “extremely difficult” to get information about patients who transferred from another health-care facility. This was, notably, in spite of a majority of the same physicians saying they had a “good sense” of how to get transfer materials and reporting that they regularly evaluate a new patient’s file when that person arrives under their care.

When the system fails, leaders in medicine suggest some tricks for improvement: better documentation, better verbal handoffs, and double-checking it all. The advice usually amounts to this: When the system fails you, be more careful. Work harder.

And so we do. But on a large scale, it breaks down. We can do every one of these steps and more, but stories like Michael’s will persist because humans are fallible, memories are fickle, and it’s an intellectually gargantuan task to distill decades of history in a few sentences.

We can cold-call retired physicians from distant hospitals and we can print rhythm strips to stuff in wallets. Something relevant will eventually go missing. Maybe not this time, and maybe not the next. But as an aggregate, this can’t be counted on as a backbone system of safety. It’s also a waste of resources to rewrite a new chart every time a patient enters a new building. I’ve seen doctors go above and beyond in every possible way and yet I’ve seen how hard it is to always get it right. It’s as engineer W. Edwards Deming said: “A bad system will beat a good person every time.”

When hospitals share information, care for everyone improves. Researchers at the University of Michigan found that when emergency rooms shared files, it was far less likely for patients to have repeat CT scans, ultrasounds, or chest X-rays. Another study, from Israel, found that sharing health data, compared to looking at only internal data, cut down on redundant hospital admissions.

Slowly but surely, hospitals are venturing into data sharing. For instance, one of the most widely used electronic health vendors, called Epic, has a “Care Everywhere” platform, which helps facilities share electronic patient records quickly and efficiently. More recently, the company upped the ante with a “Share Everywhere” tool, giving patients control to share their health data with doctors anywhere in the world. Care for veterans improved when the electronic health-record system VistA (Veterans Information Systems and Technology Architecture) included a tab for sharing data from one veterans’ hospital to the next anywhere in the country.

These are powerful steps. But many experts see the last frontier as a universal electronic health record accessible across all hospitals and patients. Such a health record would transcend institutional borders. It would mean no more creative work-arounds. No more paper faxes. No gaps.


Michael’s medical team eventually corrected his high sodium and, after two months of rehab, Leah finally took him home—with hospice services. Generally, in order to qualify for hospice, a doctor must certify that someone has six months or less to live. But sometimes when we cut out the aggressive medical care, amazing things happen. Some people live longer.

Such was the case with Michael.

On a windy afternoon in January 2018, nearly two years after I first met the Champions, I saw them once more, this time at their home. Michael was doing well, considering what he’d been through. Leah had decided to do Michael’s insulin and water herself, and she also had nursing aides come visit the home four days a week. In the end, she trusted herself most. She was the pillar of continuity Michael needed—that many others do not have.

During my visit, I told Leah that I never found out exactly where the breakdown in communication happened at the nursing facilities nearly two years ago. I retraced their steps and paid visits to both, trying to re-create their story. Along the way, I found no electronic trail of Michael’s stay. There were some records, but for discharge patients they were stored in a paper binder in a separate storage facility.

I sat with the admissions coordinator at one of the nursing facilities as she ran me through the process. How they obtain records depends on where the patient is coming from, she explained. Some places send electronic forms and some go through fax. What arrives is a standard set of papers: the physician’s discharge summary and the most recent medication list. The nursing facility doesn’t receive a full set of lab values unless the physician copies them into the discharge summary, or any notes prior to the hospitalization, such as primary-care records.

I wish finding the hole was simple: a broken fax machine; a paper chart that ended up on the wrong desk. But when the process is roundabout, requiring multiple steps and multiple work-arounds, the breakdown points multiply. Sometimes, it’s remarkable that things turn out well so much of the time, given all the places they could go wrong.

Michael had more fortitude than almost anyone I’ve met. He recovered, again and again—sometimes because of, and sometimes in spite of, his medical care. But six months after my visit, just this past July, Michael died—on his own terms. He’d had pneumonia and it showed signs of coming back. Leah asked if he wanted to treat it or have nature take its course. “His gift to me was clarity,” Leah told me later. “I didn’t have to make the decision for him.” Michael’s last words to Leah had been: “Let it go.” She brought in family and friends to say their goodbyes. Then they stopped his tube feeds.

Since then, I’ve thought back to my last visit with Michael, when Leah asked if I’d like to talk to him privately. When we were alone, I asked him: “Are you happy?”

Michael looked out the window and then turned back to me.

“Yes,” he nodded.

I nodded back, reflecting on what it was like to reconstruct the entire human narrative of Michael Champion from a handful of scattered, disconnected fragments. As I looked at him, I saw all my patients, and I thought of how to get them the informed medical care they so deeply deserve. I eagerly await the day a universal electronic health record connects the dots. But it’s not that day, so we in the health-care system must do everything in our power to deliver good care to those who trust us with it. We will continue to push papers through fax machines, to wait on hold as we cold-call those who may provide answers, and to repeat tests from scratch when we’re stalled.

We know it’s not a perfect system. We know there will be gaps. But what choice do we have?

This post appears courtesy of Undark Magazine.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.