The criminal justice system has failed to stop the homicide epidemic in America, so public health experts at the Centers for Disease Control in Atlanta have decided to treat it like a disease.

"Our first step has been to describe the problem," says sociologist James Mercy, assistant director of CDC's relatively new Violence Epidemiology Branch.

CDC's numbers paint a horrifying picture:

* Homicides accounted for 178,467 American deaths between 1970 and 1978, more than three times the U.S. losses in Vietnam. The number of homicides rose sharply in the late '60s, peaked in 1974, and leveled off at about 23,000 a year. Men were the victims 78 percent of the time.

* Homicide mostly strikes the young. For Americans between 15 and 34, it is the second leading cause of death after accidents. Suicide ranks fifth.

* Homicide strikes non-whites about 5.6 times more often than whites. It is the leading cause of death in American black men between 15 and 34.

* The United States leads the industrialized world in homicides. In 1982, nine out of every 100,000 Americans died at the hands of his fellow. In Great Britan, only one Englishman for every 100,000 died that way.

* A majority of homicide victims die at the hands of relatives, friends and acquaintances, not strangers. Many die in their homes.

* Sixty-seven percent of all homicides were committed with guns.

* Most homicides occurred on weekends.

Two years ago, a CDC report on homicide echoed FBI crime reports by concluding: "Homicides that do not occur during the commission of other crimes defy the traditional prevention and deterrence strategies of law enforcement agencies. This suggests that those acts of violence unrelated to other criminal activity are . . . in need of public health prevention and intervention strategies."

This theory led the CDC to create a special violence section in October 1983, but Dr. Mark Rosenberg, a psychiatrist and chief of the new branch, admits that colleagues have been skeptical.

"The medical community is rather slow to acknowledge that violence really is a public health problem," he says. "Some people are stuck on a disease model . . . If it's not a disease, we can't develop a vaccine for it, and if we can't develop a vaccine for it, we can't do anything about it."

But vaccines and antibiotics were not the only things that eliminated infectious diseases. Changing the disease environment had a lot to do with it.

"All the infectious diseases started to decline when people developed better housing and better nutrition and better sanitation," Rosenberg points out.

Franklin Zimring, professor of law and director of the Center for Studies in Criminal Justice at the University of Chicago, likens the CDC's approach to homicide to the federal government's efforts to lower the number of deaths on America's roads.

Attempts to make driving safer by changing the drivers' behavior failed. But the death toll did fall when the driving environment itself began to change as engineers added padded dash boards, collapsible steering wheels, breakaway lamp posts and medial barriers; the 55 mph speed limit also had an effect.

Limiting the availability of guns is the one change in the homicide environment that would make a siginificant difference, say the experts.

"We have access to weapons that turn a fight or quarrel into a killing," says Dr. Donald Lunde, a forensic psychiatrist and clinical associate professor at the Stanford University Medical Center.

Lunde, however, sees little hope for gun control. "There are enough guns around now to last another 100 years," he says. "Something that hasn't been tried is ammunition control. Ammunition does have a shelf life."

Although the CDC researchers have developed a generalized view of homicide in American, they also have found differences in specific cities like Los Angles.

"L.A. is unique in its racial and ethnic mixture," says CDC's Mercy. "Any interventions, to be effective, will have to take into account the ethnic makeup of the population."

For example, "the rate of domestic homicide, family homicide, is very low among Hispanics compared with blacks and Anglos," Mercy says. On the other hand, the homicide rate among acquaintances, especially among Hispanic gang members, was high.

Blacks, too, were most likely to be killed by acquaintances, but most homicides of non-Hispanic whites were committed by strangers.

"The reason there is that most Anglo homicide victims in L.A. are killed in the context of some other criminal event, like a robbery," Mercy says.

The CDC also has a project to help the Atlanta police department sort through the 50,000 to 60,000 calls reporting family violence each year, and identify the family fights most likely to lead to homicide.

The police now refer many families to the department's domestic crisis intervention unit, a group of social workers who direct battling families to social service agencies, counselors, and others for help. "But this team is so inundated with referrals that it's very frustrating," says CDC's Mercy. "They get burned out quick."

Rosenberg said a study conducted several years ago in Detroit and Kansas City found that in cases of family homicide, 80 percent of the families had called the police more than three times during battles preceding the fatal fight. Fifty percent had called more than five times. Other studies have shown as many as 90 percent of wives killed by their husbands had previously called the police.

But the CDC researchers want to do more than just collect data about violence and homicide, Rosenberg says. "Our goal really is to do something about it."

The CDC is working on ways to set up educational programs aimed at known high-risk groups. For example, he says, young men can be taught that "it's better to be a chicken for 10 minutes than to be dead."