The Detroit News
October 18, 1992
by Nickie McWhirter

Computer blows out smoking-related death figures with no real human facts.

I recently read that 435,000 Americans die every year from smoking-related
illnesses. That sounds like a rock-hard, irrefutable fact, and pretty scary.
How are such statistics determined? I phoned the American Lung Association's
Southfield office to find out.

No one there seemed to know. However, a friendly voice said most such
number[s] come from the National Center for Health Statistics. That's a
branch of the National Centers for Disease Control. The friendly voice
provided a phone number in New York City.

Wrong number. The New York office collects only morbidity data, I was told. I
needed mortality data. I was given another phone number to try. Wrong again.

Several bureaucratically misdirected calls later, I spoke with someone in
Statistical Resources at NCHS. He said his office collect mortality data
based on death certificates. Progress! Data is categorized by race, sex,
age,geographic location, he said, but not smoking. Never. No progress.

He suggested I phone the Office of Smoking and Health, Rockville, Md., and
provided a number. That phone had been disconnected. Was I discouraged? No!
Ultimately, and several unfruitful phone calls later, I found a government
information operator in Washington, D.C., with a relatively new phone
directory and a helpful attitude. She found a listing for the elusive Office
of Smoking and Health, in Atlanta.

Bingo! Noel Barith, public information officer, said the 435,000 figure
probably came from its computers. S&H generates lots of statistics concerning
"smoking-related" stuff, he said. It's all done using a formula programmed
into the computers. Really? Since I had already determined that no lifestyle
data on individual patients and their medical histories is ever collected,
how can the computer possibly decide deaths are smoking related? Barith
didn't know. Maybe the person who devised this computer program knows. Barith
promised to have a computer expert to return my call.

The next day, SAMMEC Operations Manager, Richard Lawton, phoned. SAMMEC, I
learned, is the name of the computer program. Its initials stand for Smoking
Attributed Morbidity, Mortality, and Economic Cost. The computer is fed raw
data and SAMMEC employs various complex mathematical formulas to determine
how many people in various age groups, locations, and heaven knows what other
categories are likely to get sick or die from what diseases and how many of
these can be assumed to be smoking related.

Assumed? This is all guess work? Sort of. Lawton confirmed that no real
people, living or dead, are studied, no doctors consulted, no environmental
factors considered.

Lawton was absolutely lyrical about SAMMEC and its capabilities, however,
provided one can feed it the appropriate SAFs. What are SAFs? "That's the
smoking attributable fraction for each disease or group of people studied,"
he said.

It sounded like handicapping horses. Lawton began to explain how to arrive at
an SAF, using an equation that reminded me a lot of Miss Foster's algebra

"Wait a minute!" I commanded. "I don't need to know that. I need to know if
the SAFs and all the rest of this procedure and program yield valid, factual
information.To know that we must know if sometime, somewhere, some human
being or human beings actually looked at records of other human beings,
smokers and nonsmokers, talked to their doctors, gathered enough information
from reality to begin to devise a mathematical formula that might be applied
to large groups of people much later, without ever needing to study those
people, and could be expected to yield true facts within a reasonable margin
of error? Who did that? Can you tell me, Mr. SAMMEC expert?"

Nice guy, Lawton, but he didn't have a clue. He said he thought the original
work concerning real people, their deaths and evidence of smoking involvement
was part of work done by a couple of epidemiologists, A.M. and D.E.
Lilienfeld. It's all in a book titled Foundations of Epidemiology, published
about 1980 by Oxford University Press, he said. SAMMEC came later, based on
the Lilienfeld's work. Maybe. He wasn't sure.

I was unable to find the book, or the Lilienfelds.

So there you have it. Research shall continue, but so far it has only
revealed that no one churning out statistics knows anything about smoking and
its relationship, if any, to diseases and death. A computer knows everything,
based on mystical formulas of unknown origin, context and reliability. Raw
data in, startling statistics out. SAMMEC speaks, truth is revealed! Oh,
brave new world.

Are there 435,000 smoking-related deaths per year in America? Maybe. I can
tell you this with absolute certainty, however: No human beings are ever
studied to find out.