Public Health Consultation

Appendix E: Differences in methods and findings between the NYSDOH/ATSDR health statistics review and Dr. David Carpenter's statistical summary

  1. What were the findings of Dr. Carpenter's one-page statistical summary?

    In the six zip code areas he analyzed, Dr. Carpenter reported higher than expected hospitalization rates for chronic bronchitis and chronic obstructive pulmonary disease (COPD) combined as well as for all forms of infectious respiratory disease (not defined in the summary). No other respiratory diseases were reported.

  2. What were the findings of our health statistics review?

    We found lower than expected rates of hospitalizations for respiratory illnesses (Acute Bronchitis, Asthma, Chronic Bronchitis, Emphysema and Total COPD* in an eight zip code area most likely impacted by AES power plant emissions. COPD (NOS)**, a component of COPD, was higher than expected but only in the lower potential exposure area ( 2 ZIP codes); the rate for COPD(NOS) was lower than expected in the moderate and higher potential exposure areas. In addition, when combined chronic bronchitis and COPD (NOS) were significantly lower than expected both in the lower exposure area and in all areas combined.

    * Total COPD = Chronic Bronchitis + Emphysema + COPD(NOS)

    ** COPD(NOS) = Form of COPD not specified by physician

  3. Why do the results of our health statistics review differ from those in Dr. Carpenter's statistical summary?

    Although the source of the health data (SPARCS hospitalization data) was the same for our health statistics review and the statistical summary done by Dr. Carpenter in 2005, there were several methodological differences between the two investigations that may have lead to the differing results.

Different study area/population:

We chose an 8 ZIP code study area based on an air modeling analysis performed by scientists at the Department in consultation with air pollution modeling specialists at the NYSDEC. The analysis took into account local weather patterns, terrain and facility characteristics (such as stack height). Wind patterns in the area are predominantly from the south and west meaning that the area most likely to be impacted by pollutants from the facility would be in the opposite direction (i.e. to the north and east of the facility) the majority of the time. After modeling the most likely area to be impacted by pollutants emitted from the facility, we further subdivided the area into higher (5 ZIP codes), moderate (1 ZIP code) and lower (2 ZIP codes) potential exposure areas based on increasing distance from the facility. A map showing the prevailing wind patterns and the ZIP codes used by each analysis is shown in Figure 1 of this appendix.

Dr. Carpenter chose 6 ZIP codes which lie predominantly to the south of the facility. We do not know the basis for this selection. Only one ZIP code (14441) was common to both analyses (See Fig. 1).

Different comparison area/population:

We chose our comparison area based on counties that had similar rural characteristics to the study area. Forty upstate counties without metropolitan areas greater than 100,000 people as defined by the 2000 census were chosen as the comparison area. This helped to control for not only urban/rural characteristics but also for race and socioeconomic status to some degree, as these areas tended to be more similar to the study area than metropolitan areas would be.

Dr. Carpenter's comparison population was chosen based on ZIP codes outside of New York City which had no hazardous wastes sites.

Neither Dr. Carpenter's approach nor ours excluded ZIP codes based on whether or not they had power plants or other major sources of air pollution. The differences in selection methods, however, could have resulted in the differences in results observed.

Different time period:

Our health statistics review used SPARCS data from 1986-2005, while the statistical summary done by Dr. Carpenter included SPARCS data from the years 1993-2000. In general, a longer time period would lead to a greater number of cases evaluated which, in turn, would lead to more stable estimates of hospitalization rates.

Different criteria for selection of individuals with respiratory conditions:

Somewhat different respiratory outcomes were evaluated, and the methodology used to select individuals with respiratory conditions was different in each analysis. In our health statistics review we used only the primary diagnosis code listed on the SPARCS record to select individuals with respiratory conditions. The primary diagnosis represents the illness for which the person was admitted to the hospital. There are 14 additional diagnosis codes listed on SPARCS records however, for conditions that co-existed with the primary diagnosis at the time of admission or developed subsequently to admission.

Dr. Carpenter's statistical summary included cases with any report of a respiratory illness in any of the 15 diagnosis categories. Thus, persons admitted to the hospital for reasons unrelated to respiratory illness could be included in the analysis. For example a person who was admitted to the hospital with a broken leg due to a car accident but also had acute bronchitis (caused by a cold) at the time of admission may be included in the analysis. Because approximately 60% of all respiratory diagnoses are for secondary diagnoses this could result in substantially different findings.

Health statistics review
Dr. Carpenter's
Statistical summary
Study area 8 ZIP codes mostly east and north of the facility 6 ZIP codes predominantly south of the facility
Comparison area 40 rural Upstate counties ZIP codes in Upstate and Long Island without a hazardous waste site
Time period 1986-2005 1993-2000
Outcomes reported Acute bronchitis; asthma; chronic bronchitis; emphysema; COPD (NOS); COPD (total) Chronic bronchitis + COPD (combined); all forms of infectious respiratory disease combined
Selection criteria Primary diagnosis Primary diagnosis or any other co-diagnoses (15 total)