Total AIDS cases by 1991: 270,000. Total deaths: 179,000. Total deaths in 1991 alone: 54,000. Number of Americans already infected: 1.5 million.

Where do these numbers come from?

Known as the Coolfont estimates, they are the Public Health Service's official prediction, issued a year ago this week at the Coolfont Conference Center in Berkeley Springs, W.Va., of how bad the AIDS epidemic will be by 1991. For the past year, politicians, health officials and journalists have been using them.

But even the scientists who compiled these figures say they are guesses based on guesses.

"A group of about 10 scientists got around the table at Coolfont, scratching our heads," recalled Meade Morgan, chief of the statistics branch of the AIDS program at the Centers for Disease Control.

"When I explain to people how we came up with these figures," Morgan said, "they say to me: 'Gee, how can you believe any of this?' "

But, he said, the estimates are checked every three months against actual cases reported and have held up so far without any major revision. They have been within 5 percent of the actual reports.

In predicting the future course of AIDS, the scientists sitting around the table at Coolfont faced two major questions: How many people are in the major risk groups for AIDS -- primarily gay men and intravenous drug users -- and how many of those people are infected with the AIDS virus?

The best available data on infection rates at that time came from studies of gay men recruited from clinics that treat sexually transmitted diseases and ofintravenous drug abusers recruited from methadone clinics. But both groups tended to represent "the high risk of the high risk," Morgan said.

To estimate the number of gay men at risk for AIDS, the Coolfont group relied on the 1948 Kinsey report on sexual habits in American men.

"There's your first problem," Morgan said. "This was a 1948 study, Kinsey didn't ask the right questions for our needs, and he used questionnaires without any scientific sampling."

But Kinsey's was still the best data. Using Kinsey's data, the Coolfont team estimated that 4 percent of white men between 16 and 55 are exclusively homosexual, another 4 percent have been exclusively homosexual for three years or less and another 10 percent have had at least as much homosexual as heterosexual experience for three years or more. It was assumed that 18 percent of the exclusively gay men are at risk and 10 percent of the others.

From National Institute on Drug Abuse data, the Coolfont team estimated that 750,000 Americans use intravenous drugs at least weekly and another 750,000 less than weekly. It assumed 30 percent of the first group and 10 percent of the second group are at risk.

"What we really wanted to know is how often homosexual men engage in receptive anal intercourse and how often intravenous drug abusers share needles. We just don't have the answer," Morgan said.

The Coolfont group also assumed that of the 20,000 hemophiliacs in the United States, 70 percent test positive for AIDS antibodies, though only about 3 percent have developed symptoms. Other groups, accounting for less than 10 percent of cases at the time, were assumed to have a negligible rate of infection.

Summing those risk estimates produced a total of between 1 million and 1.5 million Americans infected with the AIDS virus in 1986 -- a figure that Morgan said "everybody around the table was comfortable with."

The Coolfont projections assume that 20 to 30 percent of those infected in 1986 will develop AIDS by the end of 1991.

Morgan said it was "clear that the infection is still confined primarily" to the major risk groups -- gay men, intravenous drug abusers and hemophiliacs -- even though the virus can be spread heterosexually from men to women and women to men.

The doubling time for AIDS -- how long it takes for the number of reported cases to double -- has gradually lengthened from five months in 1981-82 to 14 months currently. By 1991, Morgan said, the doubling time is expected to be 24 months.

So how many Americans will be infected with the AIDS virus by 1991?

"If the question is whether there will be 20 million Americans infected by 1991, the answer is no," Morgan said.

"Ten million? Very doubtful. Five million? Possibly. But my feeling is, it could be well below that.

"We just don't know. That's the bottom line."