Do no harm

Nora Boström died in a hospital room, her arms clenched around her mother’s neck, on November 22, 2013. It was 22 days before her fourth birthday.

Nora had blonde, curly hair and a big laugh, and seemed to hate wearing pants — pictures of her as a toddler show her wiggling right out of them. Nora was also born prematurely with underdeveloped lungs. A few months before her third birthday, she underwent a small surgical procedure that placed a thin, snakelike tube running through her chest to her heart. Doctors used it to pump medicine into her bloodstream that would help her lungs grow.

The tube is called a central line catheter, and doctors insert millions of them into patients each year. Because they run straight to the heart, central lines are the fastest, most effective method of delivering often lifesaving medication. But if bacteria manages to get into the central line — when a nurse changes a dressing or injects a medication — it can quickly become a bloodstream infection. At best, these infections cause suffering for already-sick patients. At worst, they kill them.

Nora had four central line infections in her last year of life. "Every line infection just took more out of her, and more out of her, because it weakened her heart," Claire McCormack, Nora's mother, says. "It just weakened that perfect heart."

It's easy to write off Nora's story as a medical horror, a rare tragedy that struck a little girl already fighting a serious health condition. That's wrong: Federal data shows that what happened to Nora happened at least 9,997 times in 2013. Central line infections remain a leading cause of death in the American health-care system — despite research showing that they are nearly all preventable.

"If there were maybe a couple dozen of these each year, I’d shrug and say, given the size of this country, I think that’s acceptable," says Ashish Jha, a professor at the Harvard School of Public Health and practicing internist. "The fact that we have thousands? That points to a much bigger problem in health care."

Jha’s attitude represents a sea change in American medicine. Fifteen years ago, doctors saw central line infections as a horrible but unavoidable side effect of modern medicine. Inserting a foreign object into a patient's body just came with the risk of pathogens entering too, or so the thinking went. Patients contracted about half a million central line infections between 1990 and 2010.

A series of experiments in the mid-2000s changed that thinking. Researchers showed that doctors could significantly reduce, and in nearly all cases eliminate, central line infections if they followed a short safety checklist.

Vox’s Johnny Harris and Sarah Kliff traveled to California to understand why a 4-year-old girl experienced serious medical harm during her last year of life.

Central line infections fell 46 percent between 2008 and 2013, a huge success for public health. At the same time, these figures exasperate experts: Given everything we know about preventing central line infections, why do they happen at all? And why did they happen, four times, to Nora?

I. Plane crash hospitals vs. car crash hospitals

On March 24, a pilot named Andreas Lubitz took off from Barcelona flying Germanwings Flight 9525. The airplane was scheduled to land in Dusseldorf two hours later. But Flight 9525 never made it to Germany; Lubitz deliberately crashed the Airbus into the French Alps, about 100 miles northwest of the coastal city of Nice. All 144 passengers and six crew members died.

The easy response to this tragedy would have been fatalism: It’s extraordinarily rare for a pilot to commit murder-suicide in the cockpit, but if he does, there’s really no way to stop him. But that wasn’t Lufthansa’s response. The company realized that the best way to protect against a single person crashing the plane was to require every airplane to have two people in the cockpit at all times. This was already standard protocol in the United States, but not widespread in Europe. If a second person — a co-pilot or even an airline attendant — had been in the cockpit with Lubitz, he or she would have had an opportunity to intervene.

On March 27, three days after the crash, Lufthansa issued a policy change: It would require every flight to have two pilots in the cockpit instead of one. Other European airlines quickly followed suit, implementing similar protocol.

Whenever a plane crashes, airlines assume something needs to be fixed on every plane that will ever fly again

This type of investigation that results in policy change is typical for the aviation industry. Whenever a plane crashes, whether by mechanical failure or operator air, airlines and regulatory bodies immediately assume something went wrong — something that needs to be fixed on every plane that will ever fly again.

The reaction to fatal plane crashes is decidedly different from the reaction to fatal car crashes. Car accidents killed 32,719 people in 2013, about 90 people each day. But car companies don’t see each car crash as a failure that needs to be fixed. With thousands of wrecks happening each year, they don’t even have the manpower to investigate each crash. Fatal car accidents are seen as a sad but unavoidable inevitability of hundreds of millions of Americans getting behind a driver’s wheel each day.

Instead of looking at each crash as its own failure, car manufacturers look at large numbers of crashes to detect whether there are repeated failures. As Denny Gioia, a former Ford employee who worked in the company’s recall office, recently told the New Yorker’s Malcolm Gladwell, "Idiosyncrasies won’t do. Question is, do you have enough here indicating that these failures are not just one-off events?"

There is a crucial difference between the automobile and aviation industry. Car companies acknowledge the idea of a "one-off event": that some accidents are unavoidable, no matter how much work goes into prevention. In the aviation industry, however, one-off events just don’t exist. Airplane manufacturers treat each crash as potentially preventable and work backward to figure out how it could have been prevented.

A similar divide exists in modern medicine, when it comes to patient harm — especially for patient harm from central line infections. There are hospitals in the United States that view some level of central line infections as a sad but inevitable effect of putting thousands of these tubes into patients’ bodies each year. And then there are other hospitals that see each central line infection as a failure that requires investigation and better preventive techniques in the future.

In other words, there are car crash hospitals and there are plane crash hospitals.

Explore how often this harmful infection occurs at hospitals near you

II. A death — and a revolution — in Baltimore

Fifteen years ago, all American hospitals were car crash hospitals. "We told ourselves this story, at the time, that these infections were inevitable and that if you get an infection, it is what it is," says Peter Pronovost, a critical care physician at Johns Hopkins University.

A central line is a thin, snakelike tube that runs to the heart to deliver important medication. (Tyson Whiting/Vox)

But Pronovost began to doubt his profession’s fatalism in 2001, after the death of Josie King, an 18-month-old burn victim. She was recovering from the initial injury — second-degree burns covering her small body — when she contracted a central line infection, and died three days later.

King’s death prompted Pronovost to examine the research on central line infections, and he realized, to his surprise, that they could be prevented. The problem was that prevention was so complex as to seem almost impossible.

The Centers for Disease Control and Prevention, for example, had a 150-page document recommending 90 different things that research had shown to prevent central line infections.

"Most evidence reviews don’t rank which guidance is most important," he says. "They review literally all the therapies that might be beneficial, and say, ‘Here’s evidence for or against that.’"

Pronovost knew it would be impossible to ask doctors to do 90 different things all in the service of preventing infections. So he and some colleagues decided to look for items that met two criteria: highly effective and low-risk.

This helped create a simple five-item checklist that centered on obsessive, meticulous cleanliness when inserting the central line and changing the dressing: washing hands, covering yourself and the patient in sterile clothing and drapes, using the antiseptic chlorhexidine on the site, avoiding groin-area catheters (which get infected more often), and removing catheters when not needed.

Tap or click the image for a larger version. (Javier Zarracina/Vox)

Pronovost started to observe the surgical intensive care unit at his own hospital. He found doctors followed these five steps only 30 percent of the time.

"The hospital wasn't just set up for it," Pronovost says. "If you wanted to follow the checklist, you had to go eight different places to get everything you needed. For a doctor, that's time you're not spending doing something else. And the infection? It's invisible and in the future — not like that patient you need to check on right now."

So Pronovost got the surgical intensive care unit stocked with carts that had all the necessary supplies. He instructed nurses to call out doctors who were not following checklists. "That caused World War III," he says. Intensive care units often have a strict hierarchy that puts doctors at the top of the food chain. "The doctors said, 'You can’t have a nurse question me in public. It makes me look like I don’t know something.'"

Despite the initial resistance, Pronovost got quick proof that the checklist worked. Central line infections at the Johns Hopkins University surgical intensive care unit fell by 50 percent within three months of initiating the checklist. By six months, they were down 70 percent.

This was encouraging, but Johns Hopkins was only a single hospital — and it had Pronovost at the helm, pushing doctors to comply with the checklist. The next step was seeing whether the approach could scale to other hospitals across the country.

Pronovost’s work at Hopkins caught the eye of the Michigan Hospital Association, which was looking for ways to improve quality of care. In 2004, the two began partnering on a larger checklist study. Of about 100 hospitals in Michigan, 60 signed up for the experiment. They had some leeway to adjust the Pronovost checklist to adapt to their own hospital, so long as it still aligned with the original CDC guidance.

One of the doctors recruited to help with the effort was Robert Welsh, head of thoracic surgery at Beaumont Health System, a hospital chain just north of Detroit.

Welsh recalls approaching the experiment with caution. "When I got involved in the project, I had some skepticism," he says. "The mindset by most people prior to 2004 was that infection was just an accepted risk for having one of these catheters placed."

But Welsh got to work. He had his hospital change the type of antiseptic they used to clean central line sites. They covered every patient in full sterile drapes. Nurses received a four-hour training in how to insert central lines correctly — and were taught to intervene with doctors who didn’t follow protocol.

They saw results. "We started seeing zeros show up in our tracking," Welsh says. "They weren’t always zeros, but they were frequent. About 12 or 18 months in, instead of one or two infections being the routine, the routine became zero. One or two infections in our intensive care unit was really odd to see."

Welsh’s results were typical for Michigan: A landmark 2006 study of 103 intensive care units there showed that within three months of implementing Pronovost's checklist, central line infections fell 70 percent.

I asked Walsh what he thought was most important in the experiment. He told me that it was really about shifting the culture — away from one that saw central line infections as a cost of doing business and toward one that viewed each incident as an opportunity for prevention and improvement.

"We were empowering nurses to stop line insertions when protocol wasn’t followed, and we were making it clear that all the steps had to be taken," he says. "After that, it wasn’t totally a surprise when the zeros started to show up. When I saw the way communication had changed, I pretty quickly believed this was going to work."

III. "She just kind of slipped away"

After Nora died in November 2013, a family friend put together collages for her funeral. (Johnny Harris/Vox)

Born prematurely at 25 weeks, Nora Boström initially weighed 1 pound and 4 ounces. Her lungs weren’t fully developed, a condition called pulmonary hypertension. She spent the first four months of her life in the neonatal intensive care unit.

Experts say that pulmonary hypertension is a serious but survivable condition. And there’s a note in Nora’s medical records that describes her as "stable until October 2012."

Just before she turned 3, Nora started having fainting episodes. Her doctors prescribed an intravenous drug called Remodulin. Administration of the drug required doctors to insert a central line in October 2012 that Claire McCormack and her husband, Thomas Boström, would use to give her medication once every two days.

On February 13, Thomas noticed that Nora was running a fever. He brought her to Lucile Packard Children’s Hospital, where doctors quickly suspected the line had become infected. On February 14, medical records show that doctors removed the line and surgeons had to "remove a lot of pus from the central line tunnel after the line was removed." Blood tests a few days later confirmed the infection.

Claire and Thomas say they saw nurses who didn’t wash their hands before using Nora’s central line. "We also observed nurses …touching bedrails after they put their gloves on and then not changing their gloves before accessing the line," they wrote in a letter to the hospital after Nora’s death.

Nora’s first line infection caused a change in her treatment plan. To avoid future infections, doctors began delivering her medication subcutaneously (through a small needle poked into fatty tissue and controlled by a pump, similar to how diabetics inject insulin). That method had its own complications — namely, it seemed to cause Nora to faint — which resulted in further hospitalizations.

"Obviously we knew it was a huge decline, but it wasn’t until after she died, when I went back through the pictures, that I really realized how much had changed for her," says McCormack. "I could see it in the pictures. She’s sitting down all the time, versus up and running. That was the difference."

Mamma has a Band-Aid, too. (Courtesy of Claire McCormack)

From April 2013 forward, Nora spent much of her time living in the hospital. Her condition was deteriorating rapidly, with her lungs and heart becoming increasingly weak. Doctors at the hospital thought the situation had become dire enough that they recommended doing a transplant on both organs. Nora’s parents decided — against the advice of their doctors — not to go forward with a transplant because they felt the procedure would be too risky.

Nora’s last hospital admission in 2013 began the same way as her first one: with a central line infection. On November 8, she arrived at the hospital after her parents noticed that her heart rate was elevated and she had trouble breathing.

The "presumed central line infection," as it was noted in Nora’s records, led to septic shock — a life-threatening condition that happens when blood pressure drops to a dangerously low level after infection.

Nora recovered from septic shock after about a week in the hospital but then contracted a second infection, parainfluenza 1, a respiratory virus that is the leading cause of croup among children.

"I remember I was talking with one of the cardiologists, after the septic shock, about when we would go home, and making a plan," McCormack remembers. "Then all of a sudden, she just couldn’t breathe. It was at night. That hadn’t happened before. They swabbed her nose, and she had a respiratory virus."

Nora’s weakened body could not, ultimately, fight off that last infection. On November 22, she was in her hospital bed, and Claire remembers Nora asking her, "What can I do to feel better?" She wanted to play with a bucket of water in the room.

"Nora put her hands in it, and I think the effort was just too much," she says. "She just turned to me and was like, ‘Hold me.’ So I picked her up and she put her arms and legs around me, and her head kind of went back and she was sort of gasping to breathe, and she said, 'Please help me feel better.’ She just grabbed my neck like ... I mean, I can still feel it. There was so much strength. And then she basically lost consciousness.

"She just kind of slipped away."

IV. Four central line infections and two tragedies

It will likely never be known whether Nora would be alive today if the central line infections had not happened. Experts say that patients like Nora — kids with compromised immune systems and chronic conditions requiring a central line for years — are among the most difficult to keep safe when it comes to preventing central line infections.

The hospital where Nora was treated, Lucile Packard Children’s Hospital in Palo Alto, is a renowned medical center. It has spent an entire decade on the US News and World Report’s list of top-ranked children’s hospitals and, by numerous metrics, is a fantastic hospital — a place where parents want their kids to get treatment.

Johnny Harris/Vox

Lucile Packard Children’s Hospital in Palo Alto, California, where Nora was treated. (Johnny Harris/Vox)

Packard does better than average on preventing central line infections, too. The nonprofit hospital rating firm Leapfrog finds that Packard has a lower than expected rate of central line infections, given the hospital’s size and its type of patients.

The hospital's doctors have clearly put effort into this specific issue. In 2014, Packard doctors published a study in the journal Pediatrics, conducted in their own intensive care unit, that showed a dramatic reduction in central line infections when they implemented an electronic safety checklist.

This all makes the hospital’s reaction to Nora’s death surprising and disappointing. All available evidence shows that Packard did not see her central line infections as a series of plane crashes, each one warranting review and discussion. Instead, they were thought to be more like car crashes that piled up along the way.

Claire and Thomas sent the hospital a letter about a year after Nora’s death requesting the chance to meet with administrators at Packard. At the same time, they also served the hospital with an "intent to sue" notice, the first step in pursuing a malpractice case in California.

Do you work in health care? Share your story.

Vox is looking to talk to doctors, nurses, and other health professionals about how harm happens — and what can be done to prevent it.

In their letter, Claire and Thomas described what they saw in the hospital — like nurses not washing their hands before accessing Nora’s line — and suggested a meeting to discuss the issues raised in the letter.

That meeting never happened. But two months later, in a letter dated January 15, 2015, Heather Foster, a risk manager with Stanford Health Care, Packard’s parent organization, responded.

"Unfortunately, the placement of central lines is associated with a risk of infection," Foster wrote. "There is a risk of infection — in the best of circumstances — which can never entirely be eliminated."

On McCormack’s concerns about handwashing protocol, Foster wrote: "Please be assured that multiple procedural protocols are in place to promote hand washing techniques. We understand and recognize your feelings regarding Nora's care, and we apologize that you were dissatisfied with your experience at LPCH." The hospital refused to meet with Nora’s parents.

Two hospital quality experts — Hopkins’s Pronovost and Harvard’s Jha — were provided with a copy of the letter. Both said it demonstrated the exact wrong way for hospitals to respond to accusations of harm: refusing to engage with the patient or family further or to discuss the possibility that something may have gone wrong.

"The way we respond to mistakes as hospitals is as tragic as the mistakes that happened, because it leaves patients feeling alone and isolated," Pronovost says. "A lot of patients don’t want to sue. They sue out of not having a voice, and feeling not heard."

Numerous studies show, somewhat counterintuitively, that patients who receive apologies from hospitals are less likely to pursue legal action. Often, patients and their family are just looking for an acknowledgement that something bad happened. They went through something terrible, and yearn to know that their voices were heard.

"Some meetings we have with patients, we don’t get deep insights or even find something went wrong," Pronovost says. "But even then, there’s a powerful healing relationship. In any human endeavor, when someone feels wronged, the way you mend the relationship is by connecting or reconnecting."

Claire and Thomas went forward with the malpractice lawsuit against the hospital in February. Because of the ongoing litigation, a hospital spokesperson declined requests to discuss details of Nora’s case specifically, and to discuss what, more generally, the hospital does to prevent central line infections.

"Lucile Packard Children's Hospital Stanford takes patient concerns very seriously and is committed to quality and safety for all of our patients," spokeswoman Samantha Dorman said in a statement. "We care deeply about ensuring their trust and providing the highest quality of care while in our hospital, and any incident that raises a concern for patient safety is a matter of extreme importance."

"Due to patient privacy issues, as well as the fact that the matter is now in litigation, we cannot comment on the specifics of the allegations. But we can say that our entire organization is committed to collaborating with families on ways to continuously improve, and we are proud of our commitment to quality and patient safety."

Nora’s story arguably is one of two dual tragedies. The first is that the central line infections caused a small child anguish, pain, and more time in a hospital. They caused her two parents immense suffering, too.

But there’s a second tragedy to Nora’s case: that her hospital treated her infections as a series of car crashes, not warranting a meeting to discuss what went wrong — and whether it could be prevented in future patients.

"First and foremost, the hospital failed to fully use this opportunity to learn," says Harvard’s Jha, who reviewed copies of the letters sent between Nora’s parents and the hospital after Nora's death. "They failed to use the opportunity to get better." In Jha’s view, that failure renders every patient at the hospital a little worse off.

Health care doesn’t have to work that way.

V. Hope in Roseville

Roseville Medical Center is a three-hour drive north of Packard, straight up the California interstate.

Behind the scenes

Vox’s Sarah Kliff spent four months reporting this project — and this photo essay explains how it all came together.

It is, in many ways, an unremarkable, 328-bed hospital that serves suburban Sacramento. But it is a plane crash hospital.

In 2005, the hospital had 11 central line infections. This wasn't terrible; it put Roseville at just about average for the United States in infection rates. But the Pronovost research was just starting to come out, and the hospital began exploring whether it could achieve similar results with its own checklist.

"Everyone in the country was saying, 'Wait a minute, these rates have to come down,'" says Barbara Nelson, the hospital's chief nurse executive.

A leadership team at Roseville surveyed the literature; they looked at things like the Pronovost checklist to figure out what they could do better to prevent central line infections. And in 2006, they changed their procedure using their own seven-point checklist that emphasized similar infection prevention techniques as those used in Baltimore and Michigan.

They stopped letting all nurses insert central lines — with more than 900 working in the hospital, it would be nearly impossible to train them all to get it right — and instead shifted the task to an 18-nurse vascular access team that would exclusively insert the lines and monitor them for infection.

The vascular access team was formed in 2006 to implement the checklist. And for seven years, Roseville did not have a single central line infection.

Perhaps more importantly, Roseville became an airplane crash hospital. When the hospital’s streak broke in 2014 — during that summer, two dialysis patients’ catheters became infected — Roseville decided to learn what had gone wrong.

"I’m not trying to put lipstick on this," says Deborah Dix, the hospital’s oncology director, who oversees central line infection prevention efforts. "I think we were doing a wonderful job, but then we made a mistake. We did a root cause analysis that’s still ongoing. We’re going over every experience on his chart, because we’re trying to understand what we can do to minimize that."

The Roseville response mirrors what airlines do at a moment of crisis: analyze the situation and implement new policies that could prevent the same type of problem in the future.

In this case, Dix went and looked at the dressings on the catheter lines of other dialysis patients at her hospital. Nurses applied some differently than others — using different techniques, for example, or equipment.

Nurses who treat dialysis patients at Roseville are employed by a subcontractor; they aren’t part of Dix’s central line team. These contract nurses had received central line training in 2007, when the hospital began to follow a checklist, but hadn’t received any instruction since.

The central line infections caused a change in procedure: Roseville will, beginning in July, require annual competency checks for contract nurses who manage central lines.

Dix says this type of work is instrumental to eliminating central line infections at her hospital. If the hospital doesn’t treat each one as an opportunity to improve, it leaves future patients vulnerable to the same type of harm in the future.

"When you have an outcome that isn't what you wanted — if you don't say, 'Is there anything we could have done better? Is there any way we could have changed?' — people are going to get really sick, and they're going to die," Dix says.

This article is the second in Vox's year-long series on fatal medical harm. Reporting for this series is sponsored by the Association of Health Care Journalists' Reporting Fellowship on Health Care Performance and supported by the Commonwealth Fund.