نتایج پیشگیری و بهداشت اولیه: درمان های مسکونی خطرات رنگ مبتنی بر سرب و شیوع مسمومیت با سرب در دوران کودکی

In order to gain a better understanding of the effects of an investment in primary prevention on health, I investigate the impact of treatment of lead-based paint hazards in housing units (the preventive action) on childhood lead poisoning (the health outcome) at the census tract level in Chicago, IL. I use the findings from the analysis to simulate and then weigh the costs of lead interventions against the potential benefits of reducing blood lead levels in children. Childhood lead poisoning presents an interesting case study of the potential of preventive care in reducing the prevalence of a disease. There is a clear, well-defined pathway of exposure (deteriorating lead paint in older homes) and no method of secondary care that effectively mitigates the negative health effects. I find that a one-tenth percentage point increase in the proportion of older housing units that have been remediated is associated with a four-tenths percentage point reduction in the prevalence of childhood lead poisoning, an elasticity of roughly 0.5. Citywide, this is roughly 2.5 cases of lead poisoning averted for every housing unit remediated. Furthermore, I find evidence that the effect of remediations in preventing the disease has improved over time. The lower bound estimates of the benefits associated with the reduction in lead poisoning – increased expected lifetime earnings and reduced medical expenditures – are two to twenty times the estimated costs of the remediations.

Childhood lead poisoning is the second most prevalent preventable disease (after asthma) in children in the United States (Centers for Disease Control and Prevention – CDC, 2005). In general, the average blood lead level (BLL) in children has been declining over the last three decades, down 90% since 1978 (Envrionmental Protection Agency – EPA, 2005). However, levels among low-income, urban children, particularly those living in older housing, remain high (Chicago Department of Public Health – CDPH, 2004). Furthermore, recent medical studies of the effects of lead poisoning on cognitive ability in children have demonstrated negative impacts at levels previously thought to be below the threshold for concern (Koller et al., 2004). There are many factors that have contributed to the decline in the prevalence of lead poisoning over the past few decades, foremost of which are the phaseout of leaded gasoline beginning in the early 1970s and a greater awareness of the disease. Currently, the greatest source of lead exposure in children is deteriorating lead-based paint in old, poorly-maintained housing. Thus, future reductions in the prevalence of the disease in children will come from the treatment of lead paint hazards in the home. The primary goal of this paper is to investigate the role that the treatment of lead-based paint in homes has played in the declining rate of childhood lead poisoning in the US. More specifically, I intend to estimate the impact of an investment in lead remediations, a non-medical approach to primary prevention, on the prevalence of elevated blood lead levels (EBLs) in Chicago, Illinois. Using these findings, I also provide evidence that the benefits to society from remediations and the resulting reduced lead exposure in children far outweigh the costs of the necessary lead hazard treatments. Rather than measure whether individual remediations prevent cases of lead poisoning, I investigate whether the aggregate remediation efforts in a given area (US census tracts) will reduce the prevalence of childhood lead poisoning there. I use the number of housing units remediated in several different ways to reflect the “health of the housing stock” in an area over time (i.e. the degree to which the housing stock is free of lead hazards). As the investment in prevention increases and the health of the housing stock improves, children face a lower risk of exposure to lead and the prevalence of lead poisoning should decline. There are several aspects of childhood lead poisoning and its prevention that provide a useful case for studying preventive care and its impacts on health. First, there is a clear pathway with a short time horizon from exposure to illness in children (lead-based paint hazards in older, poorly-maintained housing units). Thus, it is possible to isolate the effect of prevention on a health outcome from other confounding factors. In addition, the health effects of lead are irreversible. The only effective way to combat the illness is to prevent it from occurring. Lastly the potential benefits of prevention are rather large compared to the costs of increased remediation. Chicago is an ideal location for the study because it has one of the highest rates of childhood lead poisoning as well as one of the most active lead prevention programs in the nation. Chicago has more total cases of lead poisoning in children per year than any other US city (CDPH, 2004). I observe each variable by year from 1997 to 2003 and by census tract. I begin by estimating a simple linear model of the prevalence of childhood lead poisoning controlling for various changes in census tract characteristics over the study period. The panel dataset also allows me to include year dummy variables to account for aggregate unobserved factors that drive down EBLs homogeneously across tracts over time. There are potential sources of unobserved heterogeneity that differentially impact the prevalence of EBLs in census tracts (i.e. there are “problem tracts” that have high levels of EBLs and lead remediation in homes). If this is the case, OLS estimates of the effect of remediations on EBLs will be biased, leading to a spurious positive relationship between remediation and lead poisoning. I estimate fixed effects models to account for any unobserved census tract impacts. Reverse causality between the remediation variable and the dependent variable (i.e. a fraction of remediations are ordered in homes because children living there have tested positive for lead poisoning) is another potential problem. Again, OLS estimators will tend to underestimate the true effects of remediation. To account for the endogeneity of the remediation variables, I investigate different ways to calculate the variable that will purge any reverse causality. Once I account for census tract fixed effects, a one-tenth percentage point increase in the percentage of housing units remediated (a reasonable increase given the annual changes observed in the data) is associated with a four-tenths percentage point decrease in the prevalence of childhood lead poisoning, an elasticity of roughly 0.5. Citywide, from 1997 to 2003, this effect translates to 2.5 cases averted for each additional housing unit treated. Furthermore, the negative impact of remediations on EBLs strengthens over time, suggesting that each remediation may prevent EBLs in more than one child over succeeding years. I also find that there was a sharp decline in the prevalence of EBLs from 1997 to 2003 not captured by remediations or any of the control variables. This is evidenced by the strong, negative and consistently increasing coefficients on the time dummy variables, most likely accounting for the reduction in lead from gasoline and increased awareness. Lastly, when controlling for various housing and sociodemographic characteristics in the panel analysis, the proportion of the population that is black, a significant risk factor in other studies, is no longer an important predictor of lead poisoning in children. The increase in lifetime earnings and the reduced medical care expenditures from a one-tenth percentage point increase in remediation are far greater than the corresponding costs. The lower bound estimate of the benefit-cost ratio is 2:1 while the upper bound is roughly 20:1. The largest dollar benefit by far is the increase in the discounted net present value of expected lifetime earnings from an increase in cognitive ability in children (measured as an increase in IQ). The upper bound estimates are reasonable for Chicago when compared to other studies finding large nationwide benefits from reductions in mean BLLs.

Once I account for the unobserved effects of census tracts and time, the percentage of addresses remediated in a census tract has an inverse relationship with the prevalence of EBLs in that area. Because a failure to adequately account for reverse causality and time invariant effects in “problem” tracts will provide estimates that are biased upwards toward zero for the impact of remediations, any coefficients can be interpreted as lower bounds (in magnitude) on the range of effects. Furthermore, the magnitude of the impact is feasible given the summary statistics (Table 2). A one-tenth percent increase in the percent of addresses remediated is well within the range of potential changes. The resulting four-tenths percent decrease in the prevalence (an elasticity of roughly 0.5) is also feasible given the dramatic declines in EBLs over time. Applying these results citywide, roughly 2.5 cases of lead poisoning are prevented for each housing unit remediated. Because this ratio is greater than 1:1, there is some spillover effect of remediations. This spillover effect can be attributed to several sources. Multiple children living in a housing unit should all benefit from an effective remediation. Thus, it is certainly feasible that all young siblings will observe a decline in BLLs. Although less likely to produce this large of a spillover effect, it is also possible that children visiting the home will have lower BLLs. These spillover effects are potentially larger if the lead hazards are on the outside of the home, primarily in the form of lead contaminated soil. There is also a positive spillover to future child tenants in treated housing units. Low income renters, a group with an elevated risk to lead exposure, are a highly mobile population in Chicago. If there is high degree of turnover in high risk areas within the city, the potential spillover benefit to new tenants is very large. Conversely, if poisoned children move out of a remediated home into a unit with unaddressed lead hazards, then they will continue to have an EBL. This would suggest that the benefit from a remediation is period specific for those moving frequently rather than a lifetime reduction in lead levels. However, the knowledge gained by a family (behaviors that minimize exposure and recognizing lead hazards) should minimize future exposure in new housing units. The last possible explanation is that children that were once poisoned receive multiple passing screens. This would bias the percentage of EBLs downward and the impact of remediation upwards. This is unlikely as children with multiple screens are more likely to have positive than negative screens since follow-up is ordered only in poisoned children. The choice of a unit of observation (census tract in this analysis) may influence the observed effect of lead remediations on EBLs. The benefit of a case averted might accrue in a different tract than the site of the remediation. For instance, if a child from one census tract faces reduced lead exposure because he visits a remediated home in a neighboring tract, the benefit is seen in the child’s home tract. A unit of observation that characterizes where children are most likely to spend time away from their own home, such as a neighborhood designation, might lead to more accurate estimates of the effect of remediation in a given area. Census tract borders are drawn according to population estimates and therefore, do not necessarily reflect characteristics of an area. Neighborhood boundaries should better reflect the character of an area. If children are more likely to associate with other children and family members in their neighborhood, then using neighborhood as the unit of observation will lead to more accurate estimates. Unfortunately, using Chicago neighborhoods would cut the number of observations dramatically. Similarly, an estimation strategy that considers the effect of remediation on surrounding tracts as well as the child’s home tract might provide a clearer picture of the spacial impact of remediations. This strategy relies on the assumption that children are more likely to spend time in areas close to their home. Remediation was more effective in reducing childhood lead poisoning in later years. It has been noted that the requirements for proper lead remediation have become more stringent over time. Most importantly, the regulations now require that lead dust be controlled. Additionally, anyone performing a lead remediation must be certified in lead safe work practices. These strict guidelines may be driving the increased efficacy of remediations over time. This also means that the expiration of older remediations are not having a meaningful negative impact on the “healthiness” of the housing stock. It is likely that once a housing unit has been remediated, parents and landlords work to ensure that the units remain free of lead hazards. Thus, over time, the preventive action continues to avert cases in siblings, playmates, and future tenants. The final models yielded a few unexpected results with regard to the sign and precision of the coefficients on the covariates. Even the coefficient for the percent of pre-1950 housing units, thought to be a very strong predictor of lead poisoning, was not a significant predictor of the dependent variable. This is not completely surprising as the summary statistics showed that there is not a consistent, positive linear relationship between lead poisoning and the age of the housing stock. Lead poisoning increased with the age of the housing stock but then decreased in areas with the highest percentage of older housing, perhaps reflecting the greater wealth concentrated in areas with very old housing in Chicago. The finding also provides evidence that the condition of the housing is far more important than the age, particularly in a city where the majority of housing units were built prior to 1950 and contain some amount of lead-based paint. The age variable was most likely acting as a proxy for condition in other lead studies that found large, significant effects. In addition, the negative relationships between EBLs and housing value and education and the positive relationship with children in poverty (variables added to control for gentrification over time) are counterintuitive. However, these variables also have a strong negative relationship with remediations. It is possible that informal renovations are more common in gentrifying neighborhoods. As mentioned previously, the literature has shown that home renovations increase lead exposure on average when lead-based paint is present. A positive association with increased lead exposure through home renovations could explain the surprising findings. Multicolinearty between several of the housing and population characteristics included in the model could also contribute generally to the unexpected results for the covariates. The coefficients for the explanatory variables constructed using census data should be interpreted with caution. The variation needed over time in the fixed effects models to control for changes over the study period within tracts is generated using the linear trend between the 1990 and 2000 censuses. If the trends from 1990 to 2003 are reasonable approximations of the changes in tracts from 1997 to 2003, these variables should serve as valid controls. However, to the extent that the trends deviate from a linear trend from 1990 to 2000, I am not picking up the changes in important risk factors that could influence regional changes in lead poisoning. The lack of time-varying census tract data for many of the risk factors is a limitation of the work in that the changes in population and housing variables are intended to control for the evolving nature of Chicago neighborhoods. Future releases of demographic data at the census tract level will strengthen this type of population-based, spatial analysis. The costs savings in Chicago from a small increase in remediation is potentially several hundred million dollars. While this estimate may seem high, it is a reasonable figure for Chicago given the nationwide estimates reported in the literature. Several studies have estimated that the benefit of small reductions in lead exposure nationwide could be tens of billions of dollars (Schwartz, 1994, Salkever, 1995 and Landrigan et al., 2002). One study estimated that the benefit of reduced lead exposure since 1976 has been between $100 and $300 billion (Grosse et al., 2002). In addition, the benefits of remediation will continue to increase over time. A single remediation can potentially prevent lead exposure for all children living in that home in the future. Therefore, improving the “healthiness” of the housing stock benefits all future generations.