Wednesday, December 14, 2005

Playing Fast and Loose With the Data

Last week we told you how Joseph Mangano of the Radiation and Public Health Project had brought his traveling snake oil show to the area in and around Vermont Yankee. As we've noted before, Mangano typically pops from town to town, hoping nobody traces back his trail, and all the times public health authorities have rejected his findings concerning mortality statistics and nuclear power plants.

Well, it turns out that Mangano, this time in conjunction with the Clamshell Alliance, is at it again. This time, the Seabrook Nuclear Power Plant is the target, but the charge is much the same.

From the Hampton Union (New Hampshire):

Childhood cancer deaths in the last two decades increased by 19 percent in communities surrounding Seabrook Station, according to the group awarding the nuclear power plant a Dirty Dozen award on Tuesday.

In a released statement, Paul Schramski of the Toxics Action Center in Massachusetts said the information came from a study by the Center for Disease Control (CDC) in Atlanta.

However, neither CDC spokeswoman Susan Asher nor Seabrook Station spokesman Al Griffith had any knowledge of such a study, they said.
This is a familiar tactic that anti-nuclear activists employ. Essentially, they'll take selective data from a report or study, and then use it to draw conclusions that the data don't support.

That, and hope that nobody tries to double-check your data. The last time we encountered this was last February when my former colleague, Brian Smith attended an Environmental Impact Hearing in Louisa County, Virginia near the North Anna Nuclear Power plant. The following is a direct quote from Dominion Power employee Delbert Horn, and what he found when the Blue Ridge Environmental Defense League claimed that local children were harmed by radioactive emissions from the plant:
"[Zeller] claims the data suggests these children were harmed by radioactive emissions from the plant. Mr. Zeller referenced the CDC website as his data source, so I went online myself to check out his numbers "… and I encourage all of you to do the same."

"While the Blue Ridge website says their death statistics exclude accidents, homicides, and suicides, what I saw at CDC.gov proved otherwise (Louisa County). Zeller's "Before"” numbers did correctly exclude accidents, but his "After" numbers did not exclude them. This is how Zeller's death rates are made to "“almost double."
Like I've said before: Same old story, same old song and dance.

POSTSCRIPT: For more of Mangano's nonsense from the pages of The Nation, click here and here.

Technorati tags: , , , , ,

4 comments:

Delbert Horn said...

At least they are zeroing in on specific causes of death in the Seabrook numbers. In the North Anna "study," Mangano included four deaths from menengitis, one from influenza, and one "unspecified intestional obstruction."

The sample sizes involved are so small that the statistical fluctuations are huge. There are several instances where a surrounding county had zero non-accidental deaths in a particular year, then have one in the following year, then drop back to zero.

The CDC website http://wonder.cdc.gov/mortSQL.html
even tells you in the results that the death rate extrapolated to deaths per 100,000 is "Unreliable" for such small numbers.

It's sad that Mangano and his followers manipulate of children's cause of death statistics in such a callous way.

Brian Mays said...

But Delbert, you're not using your imagination. Of course, those deaths were directly linked to the presence of the plant. After all, I'm sure that those sick patients would have recovered if it was not for all of the anxiety that they were feeling from living so close to a nuclear power plant.

And the last one ... we all know that "unspecified intestinal obstruction" is just doctor shorthand for a uranium pellet stuck in the gut. That's why they put those things in child-proof containers.

Cheers...

Louis Zeller said...

The allegations directed at Blue Ridge Environmental Defense League by Delbert Horn, a Dominion Power employee, are untrue and have no basis in fact.

At an NRC hearing Delbert Horn claimed he had found an inconsistency in our analysis. According to the official hearing transcript (Work Order No. NRC-237, pages 157-161), Horn said, “While the Blue Ridge website says the death statistics exclude accidents, homicides and suicides, what I saw at wonder.cdc.gov proved otherwise. Zeller’s before numbers did correctly exclude accidents, but his after numbers did not. This is how Lou makes these numbers appear to actually double.”

The truth is that Mr. Horn himself is guilty of careless and deliberate misuse of the facts.

At issue were the public health statistics the League used for nine Virginia counties and one city nearest to the North Anna nuclear reactors. Contrary to Horn's assertion, we employed the same methodology during statistical periods before and after the North Anna power station commenced operation. We specifically eliminated homicides, suicides, and accidents. I submitted written comments, charts and tables which detailed these findings. These data are posted on our website for all to see.

I will briefly summarize the method used by the League in our analysis of public health records.

The Centers for Disease Control compiles mortality and population counts for all U.S. counties from 1979 to 2002 in a compressed mortality database. We utilized “Mortality for 1979 - 1998 with ICD 9 codes” which allows one to select a data set for location, date, age range, and other parameters. It also allows the researcher to group the mortality data by county. In our analysis of young children after the North Anna plant began operating, we selected the following value for each parameter:

Compressed Mortality
Data for Years: 1983-1986
Location: Virginia (FIPS=51)
Ages: 1- 4 years
Race: All Races
Gender: Both Genders
Grouped by: County
Crude Rate Calculated per: 100,000
Cause of Death specified by the following ICD-9 Codes: [000.1-799.9]

One of the parameters necessary for the CDC inquiry is to select the cause or causes of death. For this, one utilizes International Classification of Disease (ICD) codes. The ICD-9 Finder lists specific causes of death as follows:

001-139.8 infectious and parasitic diseases
140-239.9 neoplasms
240-289.9 endocrine, nutritional and metabolic diseases, and immunity disorders
290-319 mental disorders
320-389.9 diseases of the nervous system and sense organs
390-459.9 diseases of the circulatory system
460-519.9 diseases of the respiratory system
520-579.9 diseases of the digestive system
580-629.9 diseases of the genitourinary system
630-676.9 complications of pregnancy, childbirth and the puerperium
680-709.9 diseases of the skin and subcutaneous tissue
710-739.9 diseases of the musculoskeletal system and connective tissue
740-759.9 congenital anomalies
760-779.9 certain conditions originating in the perinatal period
780-799.9 symptoms, signs, and ill-defined conditions
800-999 external causes of injury and poisoning

For the mortality data submitted to the Nuclear Regulatory Commission by the Blue Ridge Environmental Defense League on February 17, 2005 and in an earlier compilation posted to our website referred to by Mr. Horn, we utilized ICD-9 codes 001 - 799.9 throughout. Our methodology specifically eliminated ICD 800-999, which includes homicides, suicides, and accidents, during statistical periods both before and after criticality at North Anna.

Thus, Mr. Horn’s charge that our analysis included accidental deaths after the reactors started is bogus.

Moreover, Mr. Horn’s accusations reveal a flawed logic. For example, Mr. Horn stated that in the 4-year old cohort of our analysis there was one case of death by criminal neglect and three deaths by fire. Even if these accidental deaths had been incorrectly included in the data—which they were not—the total number of deaths would decrease but slightly, from 26 to 22. However, the number of deaths among children 1 to 4 years of age before North Anna opened was 12. So, if anyone at the hearing actually accepted Mr. Horn’s opinions, the analysis would be as follows:

12 deaths in a population of 49,637 before North Anna opened = 24.18 per 100,000

22 deaths in a population of 54,075 after North Anna opened = 40.68 per 100,000

Here we see that the mortality increase in children 1 to 4 years of age after North Anna opened would still be very high at 68%. Mr. Horn’s allegations are not only specious, they do nothing to alter the conclusion that deaths increased significantly after the nuclear reactors began operation.

In the counties nearest North Anna the death rate in children 1-4 years of age increased by 99% between 1979 and 1986 while in the remaining 126 Virginia counties and cities the death rate decreased by 8%; likewise, for children ages 5-14, local increase 72%, statewide decrease 3%. A similar pattern is observed in the death rates for children under 1 year of age and in fetal deaths (stillbirths).

The data reveal a need for further investigations into morbidity and mortality in the communities around the North Anna nuclear power station.

Delbert Horn said...

"At issue were the public health statistics the League used for nine Virginia counties and one city nearest to the North Anna nuclear reactors." Are you sure, Mr. Zeller? Check the Virginia State Map at http://quickfacts.census.gov/qfd/maps/virginia_map.html

See the line between Louisa County and Spotyslyvania County? That's where North Anna is. The study you cited on your website bredl.org/press/2004/NorthAnna_deathrates.htm
states: "Mangano found that from 1978 to 1979-81 (the first years of North Anna operation), death rates rose in the nine counties nearest the plant."

"The nine counties studied have a population of about 400,000, a number that is rapidly rising. They include Albemarle (including Charlottesville city), Culpeper, Fluvanna, Goochland, Greene, Louisa, Madison, Orange, and Spotsylvania."

I don't know how it is in North Carolina, but in Virginia, county maps are taught as a fourth grade SOL. Our 4th grade students can tell you that those are NOT the nine counties nearest North Anna.

This "study" leaves out Caroline and Hanover counties, which are both less than 10 miles from the plant. Can you tell us why these aren't included in the "nine nearest counties," Mr. Zeller?

The study mysteriously includes Green county, which is more than 30 miles away, the City of Charlottesville is more than 35 miles away, Madison, Albermarle, and Culpeper which are 20 miles away. The city of Richmond is the same distance as the city of Charlottesville, but was not included. Fredericksburg is less than 25 miles away, but is not included either.

Nothing like cherry-picking counties and cities to manipulate the children's death statistics, is there, Mr. Zeller? I know this Managano "study" cited on your website is more than a couple of years old now, but maybe you can still get your money back!