FORT SILL, Okla. -- Jessica Zeppa, five months pregnant, the wife of a soldier, showed up four times at Reynolds Army Community Hospital in Fort Sill in pain, weak, barely able to swallow and fighting a fever. The last time, she declared that she was not leaving until she could get warm.
Without reviewing her file, nurses sent her home anyway, in a wheelchair, with an appointment to see an oral surgeon to extract her wisdom teeth.
Zeppa returned the next day, in an ambulance. She was airlifted to a civilian hospital, where despite relentless efforts to save her and her baby, she suffered a miscarriage and died Oct. 22, 2010, of complications from severe sepsis, a bodywide infection. Medical experts hired by her family said later that because she was young and otherwise healthy, she most likely would have survived had the medical staff at Reynolds properly diagnosed and treated her.
"She was 21 years old," her mother, Shelley Amonett, said. "They let this happen. This is what I want to know: Why did they let it slip? Why?"
The hospital doesn't know, either.
Since 2001, the Defense Department has required military hospitals to conduct safety investigations when patients unexpectedly die or suffer severe injury. The object is to expose and fix systemic errors, often in the most routine procedures, that can have disastrous consequences for the quality of care. Yet there is no evidence of such an inquiry into Zeppa's death.
The Zeppa case is emblematic of persistent lapses in protecting patients that emerged from an examination by The New York Times of the nation's military hospitals, the hub of a sprawling medical network -- entirely separate from the scandal-plagued veterans system -- that cares for the 1.6 million active-duty service members and their families.
Internal documents obtained by The Times depict a system in which scrutiny is sporadic and avoidable errors are chronic.
As in the Zeppa case, records indicate that the mandated safety investigations often go undone: From 2011 to 2013, medical workers reported 239 unexpected deaths, but only 100 inquiries were forwarded to the Pentagon's patient-safety center, where analysts recommend how to improve care. Cases involving permanent harm often remained unexamined as well.
At the same time, by several measures considered crucial barometers of patient safety, the military system has consistently had higher than expected rates of harm and complications in two central parts of its business -- maternity care and surgery.
More than 50,000 babies are born at military hospitals each year, and they are twice as likely to be injured during delivery as newborns nationwide, the most recent statistics show. And their mothers were more likely to hemorrhage after childbirth than mothers at civilian hospitals, according to a 2012 analysis conducted for the Pentagon.
In surgery, half of the system's 16 largest hospitals had higher than expected rates of complications over a recent 12-month period, the American College of Surgeons found last year. Four of the busiest hospitals have performed poorly on that metric year after year.
Little known beyond the confines of the military community, the Pentagon's medical system has recently been pushed into the spotlight. In late May, Defense Secretary Chuck Hagel ordered a review of all military hospitals, saying he wanted to determine if they had the same problems that have shaken the veterans system.
Mr. Hagel said the review would study not just access to treatment, the focus of investigations at the veterans hospitals, but also quality of care and patient safety -- issues that The Times has been looking at, and asking the Pentagon about, for months.
Defense Department health officials say their hospitals deliver treatment that is as good as or better than civilian care, while giving military doctors and nurses the experience they may one day need on the battlefield. In interviews, they described their patient-safety system as evolving but robust, even if regulations are not always followed to the letter.
"We strive to be a perfect system, but we are not a perfect system, and we know it," said Jonathan Woodson, assistant secretary of defense for health affairs. "We must learn from our mistakes and take corrective actions to prevent them from reoccurring," Dr. Woodson added.
The Times' examination, based on Pentagon studies, court records, analyses of thousands of pages of data, and interviews with current and former military health officials and workers, indicates that the military lags behind many civilian hospital systems in protecting patients from harm. The reasons, military doctors and nurses said, are rooted in a compartmentalized system of leadership, a culture of interservice secrecy and an overall failure to make patient safety a top priority.
The investigations of unforeseen deaths or permanent harm, called root-cause analyses, are widely regarded as a centerpiece of efforts to make care safer. Asked about the military's missing inquiries, James Bagian, director of the University of Michigan's Center for Healthcare Engineering and Patient Safety, said, "If in fact unexpected deaths were reported and ignored, there would appear to be no good answer for that except that someone is sleeping at the switch."
Avoidable errors can and do occur at the best of hospitals. But the military's reports show a steady stream of the sort of mistakes that patient-safety programs are designed to prevent.
The most common errors are strikingly prosaic -- the unread file, the unheeded distress call, the doctor on one floor not talking to the doctor on another. But there are also these, sprinkled through the Pentagon's 2011 and 2012 patient-safety reports:
* A viable fetus died after a surgeon operated on the wrong part of the mother's body.
* A 41-year-old woman's healthy thyroid gland was removed because someone else's biopsy result had been recorded on her chart.
* A 54-year-old retired officer suffered acute kidney failure and permanent hearing loss after an incorrect dose of chemotherapy.
Such treatment failures are known as "never events," because they are potentially so grave -- and so preventable. They do not happen frequently. But a persistent rate of such mistakes can indicate broader patterns of slipshod care.
Malpractice suits can also be a rough indicator of risk. From 2006 to 2010, the government paid an annual average of more than $100 million in malpractice claims from surgical, maternity and neonatal care, records show. It would be paying far more if not for one salient reality of military health care: Active-duty service members are required to use military hospitals and clinics, but unlike the other patients, they may not sue. If they could, the Congressional Budget Office estimated in 2010, the military's paid claims would triple.