By Robert Root-Bernstein

The Wall Street Journal 2 Dec. 1993

Statistics of any kind are difficult to interpret, and AIDS statistics are no exception. Indeed, the latest statistics released by the Centers for Disease Control are so odd as to be virtually indecipherable. One begins to suspect the difficulties have been introduced on purpose.

On Oct. 30, the CDC released a report contending that AIDS has become the top killer of U.S. males age 25 to 44, and the fourth leading killer of women in the same age group-statistics that received wide currency in the lead-up to World AIDS Day yesterday. Although the report states clearly that part of the reason for the sudden upsurge in AIDS cases is due to a new expanded definition that went into effect a year ago, it nonetheless presents the data so that the implications of this change are far from obvious. Between January and September of 1992, 60,656 people were diagnosed with AIDS and reported to the CDC. Between January and September of this year, 85,526 new AIDS patients were reported to the CDC, an increase of 41% from the year-earlier period. One's first impression is that the epidemic continues to worsen.

The CDC report immediately cautions that the huge increase in new AIDS cases does not mean that AIDS is mushrooming. Most of the increase, it says, is due to changes in the definition of AIDS that vastly expanded the diagnostic criteria at the beginning of this year. These definitional changes were predicted to more than double the number of reported cases and, indeed, the number of cases reported under the new definition account for more than half of all those reported so far this year.

Interestingly, though, no one seems to have noticed that if the number of cases diagnosed under the new definition was to be more than twice the number found under the old definition, and AIDS has continued to grow at the same rate as before, then the total number of AIDS cases reported this year should be not 41% higher than for the same period last year but more than 100% higher. Why the discrepancy? Even odder, CDC officials claim that when the new definitional bias is taken into account, AIDS is only increasing about 3% to 5% a year, as it has been since 1990.

The definitional changes simply allowed many people who were going to get AIDS someday (according to the old definition) to be diagnosed (and treated) sooner. That is good for peoplewith AIDS, but what does it do to our ability to track the course of the epidemic?

The CDC itself admits that over half of the new cases reported this year are due to the new definition: 48,915 of the 85,526. In other words, only 36,611 of the AIDS cases reported so far this year would have qualified as AIDS cases according to the old definition. That is 20,045 fewer AIDS patients in 1993 than the 60,656 diagnosed during the same nine months in 1992. This is a drop of 33%, not an increase of 3% to 5%.

So, are AIDS cases increasing steadily at 3% to 5% a year, as the CDC claims without any apparent justification - or are they decreasing drastically by some 30%? If all the people who have been diagnosed as having AIDS according to the new definition would eventually have been diagnosed as having AIDS according to the old definition a fact the CDC itself admits- then is it not possible to see the current glut of large numbers as a form of artificial inflation designed to impress the CDC's clients just at a time when, in fact, the AIDS epidemic may have peaked and be on its way down?

The crucial information is the number of people who would have been diagnosed with AIDS according to the new definition had it been in place in 1992, or 1990, or even 1985. These data do not, and can never, exist.

Even worse, this is not the first time that the definition of AIDS has been altered to increase the number of diagnoses. The definition change implemented in 1985 increased the number of AIDS diagnoses about 2% a year over the 1984 definition; the 1987 change increased new diagnoses 30% to 40% a year; and now the new changes more than double the number. Anyone simply plotting AIDS statistics provided by the CDC year by year will be terribly misled by the resulting graph for the simple reason that the numbers represent apples one year; apples and oranges another; apples, oranges and bananas a few years later; and so forth.

As James O. Mason, assistant secretary of health and human services, told the Los Angeles Times in 1991 when the latest alterations were proposed, changing definitions "messes up the baseline for comparison from past to future" and "will make the interpretation of trends in incidence and characteristics of cases more difficult." Perhaps that is the point.

Perhaps the CDC does not want anyone to recognize what some statisticians and actuaries have been predicting for several years now; that AIDS peaked sometime between 1989 and 1992 in the U.S. and is on its way down. Indeed, according to one of the best validated principles of epidemiology - Farr's law - the epidemic should proceed at the same rate that it arose.

Think of it: Perhaps those people who were most susceptible to AIDS have already contracted the disease and it will remain within high risk groups, as the National Research council report predicted last year. Or even better, maybe safer sex campaigns, clean needle exchanges and better health care are having their effects. Is it really so unthinkable that there is good news on the AIDS front that we cannot face it when it appears?

The upshot of this lesson is a simple one. The CDC is comparing apples with apples and oranges. That it can get away with such sleights of hand only goes to show how really deeply innumeracy runs in this society; innumeracy so pervasive that not even our science reporters, our AIDS activists and our researchers seem to notice. Or is it worse, even, than not noticing. Have we reached the stage in AIDS advocacy that we will mislead in order to succeed? *