Grantee Research Project Results
Final Report: Assess the Linkage Between School-Related Environment, Children’s School Performance/Health, and Environmental Policies Through Environmental Public Health Tracking
EPA Grant Number: R834787Title: Assess the Linkage Between School-Related Environment, Children’s School Performance/Health, and Environmental Policies Through Environmental Public Health Tracking
Investigators: Lin, Shao
Institution: The State University of New York
EPA Project Officer: Hahn, Intaek
Project Period: February 1, 2011 through January 31, 2014 (Extended to January 31, 2017)
Project Amount: $500,000
RFA: Exploring Linkages Between Health Outcomes and Environmental Hazards, Exposures, and Interventions for Public Health Tracking and Risk Management (2009) RFA Text | Recipients Lists
Research Category: Human Health
Objective:
The objective of this project is to develop new and improve existing environmental public health indicators (EPHIs) related to the school environment to evaluate the linkage between school environment and children's health/performance, as well as the impact of state/local environmental policy interventions. The ultimate goals are to identify school-related EPHIs, which can be used for routine and long-term environmental public health tracking and surveillance, planning appropriate interventions, and protecting children’s health.
Summary/Accomplishments (Outputs/Outcomes):
Overall project management activities
(1) Comprehensive literature search/review: A comprehensive literaturesearch was conducted on school-based hazards, public health tracking and statistical methodologies. Over 200 peer-reviewed articles were retrieved; the most relevant studies were reviewed and summarized. Major gaps in the literature were identified, such as lack of studies that examined total environmental exposures from both schools and homes, ultrafine particles exposures, and the synergistic relationships between indoor and outdoor exposures. Methodologies on exposure assessment, indicator criteria and statistical analysis methods from literature also contributed to this study. (2) IRB approval: We have applied and obtained New York State Department of Health (NYSDOH) Institutional Review Board (IRB), New York State (NYS) Data Protection Review Board (DPRB) and State University of New York (SUNY) IRB's approvals. (3) Data request and access:Approval for data access was received from the DPRB in Year 1, but access to the NYSDOH Statewide Planning and Research Cooperative System (SPARCS) data (including hospital admission and emergency department [ED] data) took 2 years after the access approval. (4) Project management: The research team held monthly meetings to check progress of each aim, identify barriers and brainstorm alternative solutions. As the Consultant of this project, Dr. Savitz provided valuable guidance and unique insight on methodology development and project implementation.
Aim 1:Develop new and enhance existing environmental health indicators related to school settings.
During the grant period, we have developed new or enhanced the existing school environmental indicators that are associated with respiratory health and student performance from the existing state data sets. We grouped these indicators into four groups: (a) Outdoor school environment or hazards indicators, including ambient air levels of O3, fine particulate matter and sulfur dioxide for each school or school district in NYS, school proximity to heavy traffic roads, large airports, industrial facilities, and hazardous waste sites (HWS) or landfills, and potential pesticide exposure; (b) School indoor environmental indicators, representing 23 variables of biological plausibly related to respiratory health, selected from the School Building Condition Survey (BCS), such as indoor air quality (IAQ) problems identified from our previous studies in NYS school buildings, such as mold or moisture problems, poor ventilation and vermin problems; (c) Health and performance outcome indicators, including respiratory disease-related hospital admissions, respiratory disease-related ED visits, school absences, and standardized test scores; and (d) Indicators for home-based environmental hazards, such as residential proximity to heavy traffic, industrial facilities, HWS or landfills and large airports, as well as home outdoor environment, air pollutant levels in the home environment and pesticide exposures. For each indicator, we systematically evaluated its data sources/completeness, geographic coverage, temporal availability and data quality issues. Furthermore, we evaluated these indicators by using seven standardized core elements or criteria, such as scientific basis and relevance (environmental health importance, public health importance, stakeholder concern), analysis soundness and feasibility (technical capacity and practicality, data and information quality), and interpretation and utility (meaningful for public health action and policy development). The detailed methods for developing and evaluating the school environmental health indicators are demonstrated in our recently submitted paper.>
Geocoding: All public schools in NYS were geocoded and assigned a latitude and longitude using MapMarker Plus, which enabled us to assign proxy hazard levels to each school. To characterize home exposure, the residential address from each hospital admission or ED visit was also geocoded. Currently, 94 percent of statewide hospital admission cases for from 1991–2006 were geocoded to street level and 5 percent to ZIP code level. Less than 1 percent of the addresses could not be geocoded. The map of geocoded addresses was overlaid onto the map of exposure regions/sources using MapInfo.
Aim 2:Explore new statistical methodology.
One of the major challenges we encountered in school environmental research is that many school environmental factors or indicators are highly correlated with each other and will create collinearity in multivariate analysis. Our study has compared methodologies on integrating multiple exposures and explored new statistical methods to handle the challenge of collinearity to reduce the number of highly correlated explanatory variables, including principal component analysis (PCA), land-use regression and cluster analysis to assess hazard/exposure-outcome relationships.
PCA: To describe and assess variation in school outdoor variables, we performed a PCA among all these factors to identify the combination or group contribution to exposures for multiple and correlated potential exposures. The concentrations of PM2.5 and O3, traffic volume at the nearest monitoring station, distance to nearest primary airport, distance to nearest major road, HIresp, ZIP code population per square mile, and ZIP code median income were entered into the analysis as continuous variables. The 16 land use classes were recoded into a five-level categorical variable, ordered by increasing likelihood and magnitude for producing air pollution. The rural-urban commuting area (RUCA) codes were recoded into a four-level categorical variable according to their primary grouping (i.e., rural, small town, micropolitan, metropolitan). The traffic distance-volume variable was also included in the analysis. Principal components with Eigenvalues of 1 or above were selected for further analysis. For interpreting what each principle component represented, 0.30 was used as a threshold for determining each variables component-specific weight.
Land Use Regression Model: Data from the 2006 National Land Cover Database were used to estimate land use patterns within 1,000 m of air monitoring sites. Data were imported into MapInfo and polygons within the 1,000 m buffer were selected. For each polygon, the primary land use category was retained. For each monitoring site buffer, the proportion of polygons falling into the following categories were calculated: "developed, high/medium intensity," "developed, low intensity," "developed, open space," and "other." Datasets were then imported into SAS and linked by census tract and air monitoring site ID. We built a model of PM2.5 based on the following algorithm (Henderson, et al., 2007): (1) Assess the association between PM2.5 concentrations and those predictors with sub-categories using a simple linear regression; (2) Keep predictors with the highest adjusted determination coefficients (R2) in each sub-category; (3) Run a stepwise linear regression on all remaining variables; and (4) Calculate the adjusted R2 of the final model.
Cluster Analysis procedure: The cluster analysis procedure (proc varclus) is closely related to PCA and can be used as an alternative method for eliminating redundant variables. In this case, we used cluster analysis procedure as variable-reduction method. We used a total of 23 school indoor and outdoor variables. The rule of thumb to retain one or more variables from a cluster is as follows: a variable selected from each cluster should have a high correlation with its own cluster and a low correlation with the other clusters. In other words, the cluster representative must have a minimum 1-R2 [1-R2 is defined as follows: 1-R2ratio= (1-R2own cluster) / (1-R2closest cluster)].
Aim 3: Assess the relationships between school-based environment and children’s health outcomes and performance.
Design and study population: To analyze the relationship between school environmental hazard indicators and student outcome indicators, case-control and cross-sectional studies were conducted, adjusting for residential outdoor exposures, SES factors, and other confounders. The study population included all children (5–17 years old) residing in NYS (excluding New York City [NYC] due to lack of NYC BCS data) and attending NYS public schools. As data on the school indoor environment from the BCS is only available every 5 years (2005 and 2010), the study period of this component was restricted to the years proximal to the BCS (i.e., 2004–2006, and/or 2009–2011).
Exposure-outcome linkage analysis: We examined how two classes of childhood outcomes (health and performance) were related to various school hazard indicators. To examine the school hazard-health association, a multilevel case-control study was conducted to compare the ratios of various hazard indicators between the cases (hospital admissions and ED visits due to respiratory diseases) and the controls (hospital admissions due to gastrointestinal diseases and accidental falls). The control group was frequency-matched with the respiratory diseases cases by age and year of admission. We have examined residential geographic variations of the controls to ensure they are geographically representative. It is necessary to use other hospital admissions as controls because we did not have direct information on the population at risk. To assess whether the control diseases (gastrointestinal and injury hospitalizations) represent the population at risk, a sensitivity analysis was conducted by comparing the sociodemographic composition of these controls with the distribution of hospitalizations for childbirth in NYS.
Although the health outcomes were at the individual level, exposure and confounding variables were classified at four levels: individual, specific school, school district and community. Individual level confounders include age, gender, race, ethnicity, residential address, insurance, length of stay at hospitals, and costs related to admission and ED visits. Residential exposure to various hazards as indicators for outdoor home exposures was measured at the individual level. School-level indicators include school building conditions, school proximity or estimated exposure to various hazardous sources, including heavy traffic, industrial facilities/HWS, large airports or pesticides. School district-level variables include school district demographic characteristics. Community-level data at the census block group level includes wind patterns/meteorological data, hospital density and proximity, and smoking patterns. A participant's school or home within 200 or 500 m of state routes was defined as traffic exposure. Schools or homes within 1 mile or 5 miles of major commercial airports were defined as airport proximity. Furthermore, schools or homes within 1 mile or 5 miles of industrial facilities, identified from Toxic Release Inventory (TRI) database, were defined as TRI exposure and the TRI annual chemicals release information was used.
To analyze the relationship between school environment, student absenteeism, and academic test scores, a cross-sectional study was conducted, and the unit of analysis for the dependent variables was the individual school level. For this analysis, the hazard and outcome indicators were analyzed at school level. However, individual student data for absenteeism and test scores were not available, and the community level factors for health outcomes might not apply to student performance outcomes.
Potential confounders: As young age and male gender are related to an increased risk of asthma, age, gender, and race/ethnicity were controlled as individual-level confounders in the analysis. Low socioeconomic status (SES) is known to be related to respiratory diseases, school absence and test scores, and therefore also was controlled using one or more measures at the school district level, including percent of students eligible for free lunch, student racial composition, school urbanicity and the need-to-resource capacity (a measure of a district's ability to meet the needs of its students with local resources). Outdoor home environmental factors are important confounders for health outcomes, and we controlled for these factors using the same measures of hazards or exposures to outdoors listed above for schools but at individual/home level rather than schools.
Statistical analysis: An initial descriptive analysis was conducted to summarize the demographic distribution of the school-aged children by school district in NYS. Bivariate analyses were then performed to identify individual variables significantly associated with the outcome indicators. The variables significant in the bivariate analysis were included in the final multivariate model. Both crude and adjusted prevalence ratios of each hazard/exposure and their confidence intervals (95%) were calculated. A multilevel unconditional logistic regression analysis was used to assess the health outcomes at the individual level in the case-control design. For student performance outcomes, a multilevel Cox model (for cross-sectional designs) was used in which the outcome variables are the rates of school absence and test scores by school. In addition to the standard Cox model, the multilevel basic hazard function was modified by incorporating an adjustment for school district random effects, which allows us to adjust for residual variation in children's outcomes among school districts. As planned, the analysis of school condition in relation to test score has been completed. Descriptive, bivariate and multivariate analysis has been performed for the relationship between BCS and school attendance rates. In addition, the analyses of hospital admissions and emergency room visits for respiratory illnesses in relation to BCS has also been completed.
Aim 4: Evaluate the relationship between environmental policy implementation and changes in school environmental public health indicators.
We have examined the environmental or health outcome impacts of implementation of three new environmental policies, including retrofitting of school buses in certain districts, anti-bus idling regulations and the NOx SIP Call.
Bus retrofit and idling reduction policies: The cross-sectional study design was used to compare hospital admissions for respiratory conditions (including asthma, chronic bronchitis, emphysema and chronic airway obstruction) in 2005–2006 between the school districts primarily using these policies (>/= 75%) and those using these policies less (< 75%). The policy information was obtained from the support of New York State Energy Research and Development Authority (NYSERDA), from which we identified 85 school districts that retrofitted their fleet in a total of 22,690 buses. The hospital admission data was obtained from SPARCS that included 12,400 children ages 5–17 years old who were admitted to hospitals for respiratory conditions. The technology used in bus retrofitting includes diesel oxidation catalyst, Closed Crankcase Ventilation or Spiracle, and coolant heater and engine pre-heater that were applied in the idling reduction program.
NOx Budget Trading Program (NBP) on health outcomes: The U.S. Environmental Protection Agency NBP NOx policies were divided into three periods: baseline (1997–2000), partial NBP implementation (2001–2003), and full NBP implementation (2004–2006). The exposure indicator includes ambient O3 at residential area only, since children do not attend school in summertime. Maximum and average daily levels of ambient air ozone pre-/post-NBP were examined to demonstrate the change after NBP policy. Similarly, the outcome indicators, including hospital admissions and ED visits due to respiratory diseases, were also examined if there were significant changes after NBP. Performance outcomes were not used, since they may be more relevant to the indoor environment. Because of using differing time-related environmental data, temporal influences and secular trends are the strongest confounders to be controlled in the time-series analysis. Though sociodemographic factors are known risk factors for respiratory illness, they are automatically controlled for in the analyses by way of the pre- versus post-comparisons.
Conclusions:
Summary of achievements
We completed all activities planned and have met the four study aims proposed during the grant period. More specifically, we successfully recruited staff, completed a comprehensive literature search and review, obtained approvals from the NYSDOH IRB, NYSDOH DPRB, and SUNY IRB. Although we faced the challenge of delayed health data access for 2 years, we finished data cleaning and linkage in early Year 3. Other components, such as developing environmental health indicators related to school setting and developing statistical methods, were completed on schedule. Despite staff changes and funding transition delay for over 1 year, we completed the analysis of linking school environment with children's outcome and environmental policy evaluation, and still completed all tasks proposed in Year 6. In general, the goals of the project had not changed, and we had achieved our project's objective and all four aims as described above.
Summary of Findings
This study produced many outputs, including a strong team with all expertise required for this project; a comprehensive relevant literature file and summary; a dataset linking school environmental factors with children's outcomes; and a summary table describing data sources, temporal and spatial coverage, and limitations. The short-term outcomes include the comprehensive evaluations of all possible indicators, new statistical methodology on exposure assessment of school environmental exposures and final predicted models of children's outcomes; the findings and magnitudes of the effects of various school environmental factors on children's outcomes, and the evaluation results of how environmental policy change affected children's health outcomes. The intermediate outcomes include four or more publications derived from this grant; multiple presentations in national or international conferences; and informing the key stakeholders, including NYSDOH Environmental Public Health Tracking and State Education Department. The long-term outcomes include the integration of study findings to enhance the current school environmental health policies or programs, development of intervention strategies to target current school environmental health problems, and eventually minimizing the effects of school environmental hazards and improving occupants' health/performance.
- A total of 23 school environmental indicators were identified and grouped into four groups: school outdoor environment, school indoor environment, health/performance and home outdoor environment. Each indicator was given a low, medium or high rank according to seven core criteria. The data sources, geographic coverage and temporal coverage were also evaluated.
- We found that many unfavorable school building conditions, such as schools with a poor rating for ventilation/IAQ or showing signs of stains, mold or water damage, and leaks in roofs or plumbing, were significantly reduced in NYS after 2005. In general, approximately one-third of schools had overall building condition ratings that stayed bad or got worse from 2005 to 2010, and approximately two-thirds got better or stayed good. Finally, building rating staying bad or getting worse was significantly associated with having no IAQ plan in place.
- Student respiratory inpatient and emergency department admissions were associated with several environmental factors, including residence/school proximity to industrial facilities (1–5 miles), heavy traffic (within 500 meters) or airport (within 1 mile), and with renovations done in school buildings.
- We also found that poor student attendance was directly associated with school proximity to airports (within 5 miles), air intake at schools near truck delivery and low sociodemographics. Poor English or math test scores were found to be associated with ventilation problems and having more than six concerning building conditions. Lower-SES schools showed stronger associations.
- Our teacher survey found a strong association between numbers of health symptoms and classroom IAQ and classroom climate such as inability to control glare, excessively dim lighting, excessive noise from all sources, and maximum classroom CO2.
- To estimate the impacts of multiple environmental policies, we compared the changes in air pollution and respiratory admissions among children from baseline to partial or full NBP implementation. All regions showed reduction in ozone levels after NOx policy implementation (with five regions experiencing significant reductions). After pooling the regional results together, we observed a 15 percent reduction of respiratory admissions in summer after full NOx policy. Stratified analyses suggested that these hospitalizations declined the most among white/other races and those covered by self-payer insurance and by Medicaid.
Journal Articles on this Report : 5 Displayed | Download in RIS Format
Other project views: | All 22 publications | 5 publications in selected types | All 5 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Kielb C, Lin S, Muscatiello N, Hord W, Rogers-Harrington J, Healy J. Building-related health symptoms and classroom indoor air quality: a survey of school teachers in New York State. Indoor Air 2015;25(4):371-380. |
R834787 (2013) R834787 (2014) R834787 (2015) R834787 (Final) R835636 (2016) R835636 (2017) |
Exit Exit |
|
Lin S, Jones R, Munsie JP, Nayak SG, Fitzgerald EF, Hwang SA. Childhood asthma and indoor allergen exposure and sensitization in Buffalo, New York. International Journal of Hygiene and Environmental Health 2012;215(3):297-305. |
R834787 (2012) R834787 (2013) R834787 (2014) R834787 (2015) R834787 (Final) |
Exit Exit |
|
Lin S, Kielb CL, Reddy AL, Chapman BR, Hwang S-A. Comparison of indoor air quality management strategies between the school and district levels in New York State. Journal of School Health 2012;82(3):139-146. |
R834787 (2012) R834787 (2013) R834787 (2014) R834787 (2015) R834787 (Final) |
Exit Exit Exit |
|
Lin S, Jones R, Pantea C, Ozkaynak H, Rao ST, Hwang S-A, Garcia VC. Impact of NOx emissions reduction policy on hospitalizations for respiratory disease in New York State. Journal of Exposure Science & Environmental Epidemiology 2013;23(1):73-80. |
R834787 (2012) R834787 (2013) R834787 (2014) R834787 (2015) R834787 (Final) |
Exit Exit Exit |
|
Muscatiello N, McCarthy A, Kielb C, Hsu W-H, Hwang S-A, Lin S. Classroom conditions and CO2 concentrations and teacher health symptom reporting in 10 New York State schools. Indoor Air 2015;25(2):157-167. |
R834787 (2013) R834787 (2014) R834787 (2015) R834787 (Final) R835636 (2016) R835636 (2017) |
Exit Exit Exit |
Supplemental Keywords:
Children, school-related hazards, public health tracking, environmental policyProgress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.
Project Research Results
- 2015 Progress Report
- 2014 Progress Report
- 2013 Progress Report
- 2012 Progress Report
- 2011 Progress Report
- Original Abstract
5 journal articles for this project