Photograph: Charles Gullung/Getty

Wednesday, may 29, 1 p.m. it had been a brilliant spring morning, but dark clouds had rolled in from nowhere. There was no rain, but thunder crashed. Janice’s room had become shady and dark, like the inside of a cave, with fluorescent light streaming in from the hall. A nurse was feeding Janice a lunchtime bottle, holding her comfortably in a rocking chair adjacent to her crib. The one-year-old was still connected to the cardiac monitor, but other than that the scene looked almost normal, a baby with her bottle, being gently rocked.

As Janice reached the end of the bottle, her eyes suddenly rolled up into her head, and her heart stopped. The nurse, seeing me walk past the door, called out. I ran in, and seeing a flat line on the cardiac monitor and Janice’s face turning blue, I took her into my left arm, covered her nose and mouth with my mouth, and began to puff air into her lungs. I puffed gently, so as not to cause a collapsed lung—a special risk in giving CPR to a small child—and with my right hand, I pressed on her sternum. Gently but firmly I pushed once every half second, hoping that this external cardiac compression would be enough to keep her blood moving from her lungs to her brain and other organs. About 15 seconds had elapsed from the time her eyes had rolled back. I had never given mouth-to-mouth resuscitation to a living person, only to plastic practice dummies in CPR classes, but that did not matter—in my mind I could clearly see everything in slow motion. On my mental checklist, all the boxes were being checked off. Janice would be fine, I told myself.

I glanced at the nurse and nodded toward the phone, and she called the code. I began to hear the cardiac arrest announced on the overhead speakers in the hallway, a woman’s monotonous voice repeating, “Code red, burn unit,” over and over again. Thirty seconds had now elapsed since Janice’s heart stopped. I looked up and saw Hal Turnbull jogging through the doorway, a sight I knew assured our success in bringing Janice back. About to begin his fifth year of training to be a cardiothoracic surgeon, Hal had received extra training in cardiac emergencies and procedures. I considered him the best surgery resident in the department. Hal grabbed an endotracheal tube and laryngoscope from the crash cart that the head nurse had pushed in behind him, and he placed the tube in Janice’s windpipe perfectly on the first try. I connected the tube to oxygen that the nurse had readied, and within three minutes after her heart had stopped, Janice’s airway and breathing were secured. Her intravenous catheters were working, and for the next five minutes we administered drugs according to the cardiac arrest protocol for standstill: atropine, calcium, bicarbonate, epinephrine. No response.

The defibrillator was ready. Hal placed the little paddles on Janice’s chest, everyone stepped back, and he said, “Shock.” Her small body twitched all over and we watched the monitor, breathless. Nothing. More CPR, more drugs, and another round of defibrillation. Again, “Shock.” But once again no rhythm returned to her heart. This was unheard of. I had been to dozens of codes that were a mess compared with this: when the endotracheal tube didn’t go in right, the defibrillator didn’t fire, the intravenous lines blew, and the drugs were spilled on the floor. Those patients had all done better than Janice was doing now. More CPR, another round of drugs, and another, “Shock.” Nothing. The lab results were coming in every five minutes or so, and her blood gases were normal. She was ventilating well through her ET tube and exchanging gases normally. This cycle of CPR, drugs, and defibrillation went on and on, for 85 minutes, as we took turns pushing with our fingertips on her tiny sternum to keep the blood moving through her stopped heart.

At some point in that crowded, shadowy room, doubt crept in and began to grow. I was perspiring, and I realized that my heart had started to race. My mind suddenly lacked clarity; everything was blurry. Something had changed. For the past hour it had not been Janice we were coding; it had been a nameless individual whose heart had stopped. Now it was all different, as I realized for the first time, This is Janice we are working on, and we may not get her back. Today, remembering clearly, I wonder if at that moment we had already lost her.

The slow motion that had enveloped me during the first hour, when it seemed like I was in control, had dissolved. I was now in real time. I felt exposed, no longer insulated by the training that renders horrible events into a sterile, intellectual exercise. Reality was fast and awful. The muscles in my chest, arms, and thighs began to quiver, and my knees felt loose, like I was falling.

Janice’s heart failed to resume beating, and Hal decided to insert a pacemaker, a sterile wire that could be introduced into the jugular vein and advanced down into her heart. The other end of the wire would be connected to a bedside generator that would produce an electrical current to stimulate her heart. I began unwrapping the sterile package and setting out the instruments as he put on his gown and gloves. By now I had ceased working mechanically. Alert, anxious, and tremulous, I knew in my own heart that this was Janice’s last chance. I was also afraid because bedside pacemakers were newly invented for such cases. I had seen them used in a few other codes, and most of the time the device could not be inserted properly, and the results were poor. But I had never been to a code with Hal Turnbull, and, miraculously, he positioned the pacemaker in her right atrium on the first try. “This is it,” said the little cheerleader in my mind, “just the break we need to turn this mess around. Janice will be fine now.”

I turned on the device and watched the cardiac monitor. On cue, the pacemaker began to fire and the electrical signal it generated appeared on the monitor, but there was no heartbeat following it. What was wrong? Every action we had taken in this code had gone by the book and perfectly, but Janice’s heart would not beat. We had arrived at the instant she needed us most, we had breathed for her and pressed on her lifeless chest, but we failed to will her heart back to its rhythmic, contractile purpose. She was dead.

I did not speak the words; Hal did. “It’s over. She’s gone.” I did not look back, and I never saw Janice again.

Dr. Brett Giroir, a pediatric intensivist based at Dallas’s Southwestern Medical Center, has noted that codes in children are inevitably unforgettable. The memories are “surprisingly visual and graphic, starting with the mass difference between the 15 or 20 adult doctors and nurses all surrounding and working on a 22-pound kid. That’s a ratio of 3,000 pounds to 22 pounds,” he says. “You just keep going and going, hoping that after all this the kid will come back and show you some physiological stability, any sign of normalcy.” I could not talk about Janice’s case with anyone for months, and I developed recurring dreams of her failed code. Unlike the real event, in which my brain had given me some protection and distance from the tragedy of her death, these dreams found me completely unprotected. From the outset of every dream I knew she would die. I frequently awoke yelling orders for drugs or saying, “Shock,” perspiring, sad and angry, breathing fast. Since that time I have spoken to countless physicians and nurses about their memories of patients whom they lost. The pediatric cases leave the deepest scars.

Tuesday, June 18. i received a manila envelope in the mail containing the report of Janice’s autopsy. We had been required by law to refer her case to the New York medical examiner because she had died after an accidental burn—she’d been scalded on her kitchen floor by cooking water. I had hoped that the report would contain answers to the major question of her case: Why did she die? Unfortunately, it contained no answers. Instead, it noted in simple, dry, pathological prose that her organs “appeared normal, there was no evidence of significant tissue damage, inflammation, pulmonary embolism, heart disease, or infection.”

The cause of death was listed as “cardiopulmonary arrest,” a phrase I knew the office had lifted from what I had written as the cause on her death certificate. Nothing was found to explain why she died. After 20 years and a research career spent looking, I have a good idea, but I still do not know with certainty.

Kevin J. Tracey, MD, ’79, is director and CEO of the Feinstein Institute for Medical Research and vice president for research at the North Shore–LIJ Health System, in Manhasset, New York. His essay is drawn from Fatal Sequence: The Killer Within, an explanation, for the lay reader, of severe sepsis. Copyright © 2005 by Kevin J. Tracey, reprinted by permission of Dana Press. The book may be ordered at a discount from the BC Bookstore.