It is a typical event in a hospital intensive care unit, a place where the typical patient receives $3,000 a day worth of doctoring, nursing, intubating and medicating.

A doctor stands outside the room of a gravely ill young man, an AIDS patient at George Washington University Medical Center, and says: "He's not getting better. I think we're going to have to speak to his family again."

The conversation will be about moving him out of this unit, which can no longer help him, into a less intense, less costly ward, where -- as is now inevitable -- he will die.

This kind of decision-making is called "resource allocation at the bedside." Another word for it might be "rationing," but sensible and even humane rationing, or so the doctors here intend.

The director of this unit is William Knaus, a leading figure in attempting to make the process both rational and humane. "Citizens without health insurance, a growing number of AIDS patients, an aging population, expensive new treatments and a growing resistance to spending more of the nation's resources on medical care -- all these have made it clear that hospital care must be rationed," Knaus says. "The only real question is how, by your ability to pay -- our current approach in this country -- or your ability to benefit?

"You should get treatment because you can benefit, not because you can pay. Because the ICU is the most expensive part of the hospital, this is the place we should start."

Much of what ICUs, and all doctors, do is irrational, based on their best judgments, habits, impressions, but too often, on incomplete knowledge.

That picture that we've seen in TV dramas? Doctors and nurses racing down the hospital corridor with their Code Blue cart, about to apply its electrodes to "save" someone whose heart has stopped.

Of all patients so treated, only about 10 percent ever leave the hospital alive. Of 503 such patients over age 70 treated at Boston hospitals, 112 "came to" but only 19 went home.

"A lot of our treatment produces a very small chance of survival," Knaus says. "In many cases, intrusive and complicated machinery is wheeled in to keep the vital signs going, to give treatment of no benefit and tremendous cost, depriving others of treatment while dignity disappears.

"Some high-cost patients are worth every cent. Without the sophisticated care they could not recover. Some 85-year-olds can benefit from our most advanced technology. But some 65-year-olds cannot. Some younger people cannot.

"We have to learn the difference."

What the ICU team, patients -- if they are conscious -- and many families are also trying to do is learn when to let go.

Wednesday, 8:30 a.m., daily rounds on 4 North, G.W.'s ICU.

A knot of 10 doctors and nurses -- the critical care team -- stands in a corridor. They discuss Patient A., 60, lungs and heart malfunctioning, a machine doing his breathing. A resident, a short young woman with long brown hair, says, "He is much more alert today. He said, 'Good morning, sweetheart.' "

"You may be right," says Brian McGrath, doctor in charge, stocky, brown hair, red moustache. "He does have a chance. But . . ." Treatment is to continue. After several days, the patient dies.

Next, Patient B., the 28-year-old with AIDS, tall, handsome and, to the untrained eye, amazingly healthy looking. But he has pneumonia. A tube in his windpipe leads to the respirator that breathes for him, and a forest of tubes and catheters protrude from his nose, neck, wrist and bladder.

He is sedated, because if he were fully awake and became irrational, he might tear out his tubes, a not uncommon occurrence. He is getting 10 drugs, three continuously. He is dying.

"He's young, which is the problem," McGrath says. "It's new for us to see young adults get this sick and die. It's always more difficult to reconcile ourselves to a young person's death. But as his prognosis gets worse and worse, we may start limiting his therapy."

The question is: Will the family agree?

"Once in a while," Knaus reports, "a family says, 'Keep on treating. It's in God's hands.' We usually abide by their wishes, though we are not obliged to give futile treatment.

"Often, however, it's just a matter of talking to them. As they see over time that the loved one is not going to get better, and complications are developing from the technology, very often their initial reluctance changes."

The ICU team moves to a conference room to discuss some past cases. Including Patient C., a 26-year-old woman with a severe, generalized infection. She also had to be connected to a breathing device.

One night, she pulled out her breathing tube. She was revived, if a heartbeat alone is revival, but she never woke up. She was kept on artificial breathing, kidney dialysis and other artifices for two weeks, still in deep coma. Her family still wanted "everything done."

Then, said her doctor, Roger Denny, "everything began getting unstable, and I called the family and talked about withholding therapy. I think they gradually started to see we weren't helping her."

The doctors finally disconnected her breathing tube. She died within 20 minutes.

Then there was Patient D., a woman of 24 with sickle cell disease. She was in sickle crisis, where misshaped cells clogged her blood vessels. She too was on a respirator. She had a severe generalized infection. Her life was in the balance.

But she left the hospital -- well -- after three weeks. The decision to treat her -- the respirator, the fluids, the drugs -- paid off.

To gauge the payoffs more accurately, Knaus and his colleagues have been working with other medical centers to develop a computerized system called APACHE for "Acute Physiology and Chronic Health Evaluation." It adds up every facet of a patient's condition and, taking account of the experiences of thousands of patients, predicts the chance of survival.

The evaluations also consider the patient's or family's wishes, whether to cling to life to the last gasp or to "let go" when further care is futile and only torture reigns. Does this reduce a patient to an impersonal number, a victim of a computer-dictated probability?

"No," Knaus says. "The computer can't think, but it can remember thousands of patients and what happened to them and tell us. It can do so to support, not make, our decisions. Its predictions, we find, are more accurate than the individual physician's. From a medical viewpoint -- if we're to make the best use of expensive resources -- we need this capability. This can be an alternative to using age to ration, to saying like {ethicist} Dan Callahan, '85 is it,' or something like it."

It is also needed, he says, from a moral and ethical viewpoint -- to dignify, not depersonalize, to give a patient the best chance for life but not the best chance for a living death.

"It's important to understand our motives," Knaus insists. "We cannot help but consider the cost of our care, the fact that resources are limited. But we'd still be motivated if we had all the money in the world.

"We recognize now that, with our technology, our technology that prevents death, we can also produce outcomes that are worse than death."

It is here that doctors like this -- doctors trying to balance rationality and humanity, economy and compassion -- need help.

Yes, attitudes have changed. "But we don't have a consensus yet on when we can deny care, a consensus on the point where treatment will have no benefit," Knaus believes. "We're still struggling to acknowledge that medicine's goal can no longer be limited to extending life but also involves determining when death is the outcome of choice.

"Living wills aren't going to be the solution. People don't want to face their own mortality. We have to come to a reasonable societal solution about the rituals associated with death."

Then both reasonable economy and humanity might reign.

Next Week: Getting medical care worth the money.