Lead Astray in Ohio: Methodology
Calculations of lead poisoned children in this report were based upon the results of the Third National Health and Nutrition Examination Survey (Phase 2, 1991-1994), as described in the Centers for Disease Control and Prevention's "Update: Blood Lead Levels - United States, 1991-1994" in the Morbidity and Mortality Weekly Report (February 21, 1997, Vol. 46, No. 7). Prevalence rates were adjusted to account for findings reported in the Morbidity and Morality Weekly Report (Sept. 12, 2003, Vol. 52, No. SS-10) "Surveillance for Elevated Blood Lead Levels Among Children - United States, 1997-2001." Population, income, and housing statistics were acquired from the United States Census 2000, Summary File 3.
Step 1: Determining Populations
According to the study discussed in the CDC's reports, the best indicators for lead poisoning risk of children are income level and age of housing, with percentages of poisoned children increasing as income gets lower and housing gets older. The study measured income in Poverty to Income Ratios (PIR), which is the ratio of total family income to the poverty threshold. Low income was defined as PIR <= 1.300, middle income as PIR 1.301-3.500, and high income as PIR >=3.501. The study also used three housing ranges, which roughly correspond to periods in which household lead usage changed. Old housing was built before 1946, medium-aged houses were built between 1946 and 1973, and new housing was built after 1973.
The primary geographic unit for our calculations is the census tract, which is the smallest area for which there is population and housing information available with the necessary level of detail. To match the categories in the study, we divided the total number of children in a census tract into categories based upon the percentages of housing units and children that fall into their respective ranges. This gives us nine population groups to use in our calculations.
Step 2: Compute Estimate of Lead Poisoned Children
The CDC report provides percentages of children aged 1-5 years which have blood lead levels greater than or equal to 10 µg/dL (the level at which they consider a child to be poisoned) for each combination of income and housing populations. Since the 2003 CDC paper reports that national prevalence of lead poisoning has dropped to a national average of 2.2% in 2000, we decreased the risk percentages by 30% from the levels reported in the 1997 CDC report. This adjustment ensured that we did not overestimate the number of lead poisoned children in Ohio. The risk percentages we used are in the following table:
|Before 1946||1946 to 1973||After 1973|
Percentage of Children with Lead Poisoning in the Nine Population Categories
To determine the number of children with lead poisoning in a census tract, we multiplied the number of children in each of the nine populations by their respective prevalence percentage, and added the results together.
Step 3: Computing State, County, and Neighborhood Estimates
The county and state estimates are simply the aggregate of poisoned children computed by census tract. To allow people to determine the risk of lead poisoning in their immediate vicinity, we also computed estimates at the Census Block Group, or "Neighborhood", scale. To calculate these estimates we distributed the number of poisoned children in a census tract into its component block groups using the more generalized income and housing census data available for that summary level.
Using data released by the State of Ohio Department of Health we determined the exact number of medicaid eligible children which were not screened. To calculate how many of these are estimated to be lead poisoned we used a methodology very similar to the one used for the census tract estimations. Because all Medicaid eligible children should fall into the low-income range, we divided them into three populations based on the housing percentages in the county. We applied the same percentages as above to the populations, and added together these results to get the estimated number of medicaid eligible children with lead poisoned children who were not identified.
Zip Code Hot Spots
While the census does provide information based on Zip Code Tabulation Areas, these regions do not correspond exactly with the Postal Zip Codes which people would be using to identify the risk for an area. Therefore, we could not provide estimates of risk by percentage of the population or number of children poisoned within a zip code. Instead, our list of Zip Code Hot Spots contains the Postal Zip Codes that contain one or more High Risk (>5% of children poisoned) Neighborhoods.
Sources of Uncertainty and Underestimation
There are several known sources which may have produced errors in these calculations, most of which result in a more conservative estimate of poisoned children. The first is the original CDC report, which was conducted with the relatively small sample size of 2,392 children, leading to a broad confidence interval. We encourage the CDC to conduct further research on the prevalence of lead poisoning to yield more reliable estimates.
Although the CDC study used a PIR value of 3.5 to divide the middle and high income populations, census data has only been released with PIR ranges up to 2.0. This means that we were forced to group many children into the high income population who in reality fall into the medium income group, and therefore have a higher risk of lead poisoning.
The largest source of underestimation was our assumption that the state-wide prevalence of lead poisoning in Ohio was the same as the national average of 2.2%. This most likely results in a very conservative result since the 2003 CDC paper reports that Ohio has the third highest number of pre-1950 housing units in the country, and therefore most likely has rates of lead poisoning that far exceed the national average. Testing rates support this, since in 2001, 6.47% of children tested in Ohio were found to have lead poisoning.