Over the years, DFACS officials have said most of these children's deaths could not have been prevented.

But the Journal-Constitution review shows that many of these children's lives were roadmaps to disaster. Whether they died directly from abuse or from accidents or medical causes, they did not need to die. Had caseworkers been trained to see the danger signs, responded more aggressively to initial complaints, had smaller caseloads and been held accountable for their work, many children's lives might have been spared. Had more people and professionals been drawn into their lives at an earlier stage, rather than shut out by confidentiality rules, children may have lived.

State Human Resources Commissioner Audrey Horne and Michael Thurmond, former DFACS commissioner, cite the final cause of the children's deaths in suggesting why their agency could do little to prevent them.

Thurmond, who fought to keep the dead children's records closed, argued there had been no increase in children's deaths in recent years. He said that asking how and why Georgia children died after coming under the government's protection was chasing "wolves that aren't there." Among those who had died, he said, "all but a few" had died from natural causes.

But an in-depth review of the 513 children's deaths shows that almost half died from unnatural causes, with 1 in 6 children dying from abuse, homicides, gun accidents or suicides.

At least nine children killed themselves after being left in homes where they suffered repeated abuse. One 16-year-old had been reported since he was 6 as a victim of severe cruelty by his stepfather, who was tried and acquitted of a child cruelty charge. The teen eventually hanged himself, leaving a suicide note written on the back of a school award saying his stepfather had finally won. "I'M DEAD," he wrote in large letters.

Horne and Thurmond are right that a large number -- almost half of these children -- died from what have been characterized in the records as medical or natural causes. Yet many of these so-called natural deaths were preventable. And a number may not have been natural at all.

Counted among the 237 medical-related deaths are two girls who were beaten as young children, rendered brain-damaged, then placed in institutions where they died years later from medical causes that stemmed from their early abuse. One died from a seizure; the other from acute respiratory failure.

Also counted among these natural deaths are babies who died because they were born prematurely. But among these 29 tiny newborns, almost half had cocaine in their systems at birth, meaning their mothers' illicit drug use during pregnancy may have led to their early births, and hence their deaths.

State officials also are correct that a minority of these children died directly from abuse. Still, in the last three years, at least 45 Georgia infants and young children were killed by adults entrusted with their care. And there may be many more.

Nationally, child abuse and neglect deaths are undercounted by nearly 60 percent, with many mislabeled as natural or accidental, a new study has concluded. Published in August in the Journal of the American Medical Association, the research was based on North Carolina death certificates.

Deaths raised suspicions

Georgia's child abuse records suggest similar miscounting occurs here.

A number of Georgia children's deaths clearly raised authorities' suspicions. When 4-month-old Toni Beauford died in 1996, officials at first attributed her death to sudden infant death syndrome, or SIDS, even though the infant had previously been diagnosed as having "failure to thrive." An autopsy later found the infant had died from an overdose of cocaine. Her mother eventually was sentenced to two years in prison for involuntary manslaughter.

That death was changed from a natural death to a homicide. But many others were never resolved to authorities' satisfaction.

Dr. Kris Sperry, the state's chief medical examiner for the Georgia Bureau of Investigation, said SIDS -- a diagnosis when other explanations for the sudden death of an infant are ruled out -- presents a particular challenge.

When two triplets died unexpectedly within three months of one another, the medical examiner noted on the autopsy, "SIDS vs. homicide?" Despite lengthy investigations by law enforcement, both deaths were eventually called "undetermined."

"The problem is you can suffocate a baby, and it is utterly indistinguishable from SIDS," Sperry said. "From a pathology standpoint, it is not possible to differentiate SIDS from an asphyxial death. We, as medical examiners, have a big wall we run into."

That means that in addition to the autopsy, a sophisticated investigation is critical, Sperry said. The child's history, including whether the child's family has ever been suspected of abuse or neglect, can tip the balance when determining the manner of death.

"These are judgment calls, clearly," Sperry said.

In addition to the questionable labeling of some natural deaths, some accidental deaths may not be as accidental as they seem. More than a quarter of the children known to child welfare agencies died in car wrecks, drownings, fires or other accidents. But a number of those accidents stemmed from their parents' neglect.

"Bad outcomes don't happen randomly in just any family," said Dr. Randell Alexander, director of the Center for Child Abuse at Morehouse School of Medicine. "Even the so-called accidental deaths are overrepresented in families where there is neglect or abuse."

Experts say the purpose of studying children's deaths is to learn how to prevent them. All of Georgia's counties are supposed to have teams of professionals that regularly review children's deaths considered unexpected or unexplained. But more than a third of the counties disregard the law and don't review all the deaths they're supposed to. Even when they conduct the reviews, their findings are withheld from the public under confidentiality laws, leaving little opportunity for the public to support or even discuss laws, policies and programs to save children's lives.

In Georgia, children die, sometimes under suspicious circumstances, and because of confidentiality laws, only a few people know. Among the 513 children who died after coming to the state's attention, few were ever brought to public light. And when they were, the public often didn't get the full story.

Risk factor: Boyfriends

Lisa Caruso of Cherokee County was a dark-haired beauty of a 4-year-old who spent hours on her family's swingset before her death in 1996. The case received ample media coverage, and the child's mother is serving an 11-year prison sentence after pleading guilty to beating her daughter to death.

But left out of the coverage were key facts of Lisa's life that spoke to how the state protects children.

Until now, there was no way of knowing that six months before her death, Lisa had been in Scottish Rite Children's Medical Center with a black eye and swollen face. DFACS investigated a report of possible child abuse.

Nor did the public know that at the time of her injuries, Lisa was not the Carusos' daughter. Rather, the Carusos were trying to adopt her. Despite the child's suspicious injuries, DFACS approved the adoption. Three months later, Lisa's new mother killed her.

The analysis of children's deaths shows that common themes run through them. Among the Georgia children who died after coming to the state's attention, more than half had mothers who were teens when they had their first child. A recent U.S. study shows that babies born to teen moms under 17 are 10 times more likely to be killed than those born to mothers over 25.

Another risk factor is single mothers' boyfriends, who commit 27 times more child abuse than would be expected given the time they spend caring for children, a national study found.

This newspaper's analysis found at least 12 children who died in the last three years at the hands of their mothers' boyfriends. In some of their cases, professionals such as doctor, teacher or police, had warned DFACS that the boyfriend had a violent past or the child had previously been beaten by the man or was afraid of him.

Serious mental illness of parents also represents a risk factor. Among the children already known by child welfare agencies, at least 45 died whose mothers had been diagnosed with mental illness.

Examples include a 22-month-old who was returned by the courts to his suicidal mother six weeks before his death. The woman, who had a history of emotional problems and was under the care of a county mental health department, had asked police to take her children.

Initially, authorities believed the baby had choked to death. But an autopsy showed he'd been strangled, and the death was changed to a homicide. Law enforcement investigated but could not identify the person who had killed the baby.

Overall, the causes of the majority of these children's deaths are the repetitive ones -- the deaths that say something about the secrets of children's lives, about the threads that weave together avoidable tragedies.

They are the deaths that point to possible solutions for saving other children's lives.