Studies Show ... Cluster or Clutter?

by David Schoenfeld with Margaret Wahl on Sat, 2003-11-01 10:37

Part 2 in a series

Support Group

"Psychiatrists report massive numbers of people with anxiety following the attacks of Sept. 11."

"Toxic fumes may have led to cluster of 40 with ALS at Texas military base."

One kind of news that gets a lot of people’s attention is a location or activity that seems to be associated with something frightening. Known as a cluster, these are geographic or occupational concentrations of diseases, UFO sightings or other phenomena.

But not everything that looks like a meaningful cluster is one. Statisticians call the kind of attention that often puts clusters on the map an ascertainment bias, which means that the way the group was identified ("ascertained") in the first place biases the finding.

Are psychiatrists’ records the best way to estimate how widespread anxiety is in a particular population? Or are psychiatrists seeing ("ascertaining") what’s actually only a tiny percentage of the population?

And were the 40 ALS cases noted because someone who worked on the base and later developed ALS drew special attention to this workplace?

Although 40 cases of ALS among former employees of one military base sounds like a lot, it may not be. Important questions are: How many former employees were there in the group who didn’t get ALS? And, what’s the expected number of cases in this age group?

Even if more former base employees than expected developed ALS, this could still be a chance finding. This ex-employee group is actually one of many potential groups of former employees (for instance, ex-employees of other companies or institutions) that could have been examined but weren’t.

Chance alone can lead to finding that one ex-worker group has a higher rate of ALS than another ex-worker group, just as chance alone can lead to an excess (cluster) of "heads" or "tails" in one series of 10 coin tosses among many coin toss series.

If the investigators don’t have a specific factor in mind that goes with the cluster — a segregating factor — then it’s particularly hard to even begin to test the cluster’s meaning.

To find out whether service on the base really had anything to do with ALS development, there would have to be at least a hypothesis about a causative agent — for instance, a specific chemical toxin or high-altitude flight duty on the base.

Then you’d have to look around the country at retirees from other bases to see whether people who were similarly exposed to the proposed factor harbored a similar number of ALS cases.

Only then could you begin to separate clusters from clutter.

No votes yet
Related Articles
MDA cannot respond to questions asked in the comments field. For help with questions, contact your local MDA office or clinic or email See comment policy