CoChat
Sign Up
Cited by 83Air Quality and Health ImpactsAtmospheric chemistry and aerosolsIndoor Air Quality and Microbial Exposure

Potential confounding of particulate matter on the short-term association between ozone and mortality in multisite time-series studies

Open access
Bell, Michelle L., Kim, Jee Young, Dominici, Francesca·2007·Environmental Health Perspectives
DOI 10.1289/ehp.10108PMC PMC2072830
Open this paper in CoChat — free accountOpen PDF

Abstract

Background: A critical question regarding the association between short-term exposure to ozone and mortality is the extent to which this relationship is confounded by ambient exposure to particles. Objectives: We investigated whether particulate matter < 10 and < 2.5 μm in aerodynamic diameter (PM10 and PM2.5) is a confounder of the ozone and mortality association using data for 98 U.S. urban communities from 1987 to 2000. Methods: We a) estimated correlations between daily ozone and daily PM concentrations stratified by ozone or PM levels; b) included PM as a covariate in time-series models; and c) included PM as a covariate as in d), but within a subset approach considering only days with ozone below a specified value. Results: Analysis was hindered by data availability. In the 93 communities with PM10 data, only 25.0% of study days had data on both ozone and PM10. In the 91 communities with PM2.5 data, only 9.2% of days in the study period had data on ozone and PM2.5. Neither PM measure was highly correlated with ozone at any level of ozone or PM. National and community-specific effect estimates of the short-term effects of ozone on mortality were robust to inclusion of PM10 or PM2.5 in time-series models. The robustness remains even at low ozone levels (< 10 ppb) using a subset approach. Conclusions: Results provide evidence that neither PM10 nor PM2.5 is a likely confounder of observed ozone and mortality relationships. Further investigation is needed to investigate potential confounding of the short-term effects of ozone on mortality by PM chemical composition.

AI snapshot

Written from the paper’s full text.

TL;DR

This study tests whether PM10 or PM2.5 confounds the short-term association between ozone and mortality in 98 U.S. urban communities from 1987 to 2000. It finds neither PM measure is a likely confounder, though data availability severely limits joint ozone–PM analysis.

Key findings

  • Ozone was not highly correlated with PM10 or PM2.5 within strata of ozone or PM concentrations; Pearson correlations ranged from < 0.00 to 0.22 for PM10 and −0.17 to 0.25 for PM2.5, with similarly low seasonally detrended correlations.
  • National and community-specific ozone–mortality effect estimates were robust to inclusion of PM10 or PM2.5 in time-series models, and robustness remained at low ozone levels (< 10 ppb) using a subset approach.
  • Data availability was limited: among 93 communities with PM10 data, only 25.0% of study days had both ozone and PM10 data; among 91 communities with PM2.5 data, only 9.2% of study days had both ozone and PM2.5 data.
  • PM adjustment widened the 95% posterior intervals dramatically and statistical significance was lost, likely because of greatly decreased sample size.

Methods

  • Time-series analysis of daily weather, pollution, and nonaccidental mortality data for 98 U.S. urban communities over 14 years (1987–2000), using NMMAPS-processed data; models estimated the log expected daily mortality rate against same- and previous-day 24-hr ozone with day-of-week and natural cubic spline covariates for temperature, dew point, and time, then combined community estimates via Bayesian hierarchical modeling.
  • Confounding was assessed by: (a) correlations between daily ozone and PM10/PM2.5 stratified by ozone or PM levels, season, region, detrended data, alternative ozone metrics, and lag structures; (b) inclusion of PM10 or PM2.5 at lag 1 day as a covariate in community-specific time-series models; and (c) a subset approach fitting models only for days with ozone below cutoffs from 10 to 60 ppb, with PM10 adjustment.

Limitations

  • The paper states that data availability severely limits joint analysis of ozone and PM, especially for PM2.5; PM2.5 adjustment was not applied in the subset approach because of sample size constraints, and only 62 communities had sufficient ozone and PM2.5 data for community-specific time-series analysis.
  • The paper states that the work uses total PM mass by size distribution and does not account for different particle chemical compositions; further investigation of confounding by PM chemical composition is needed.
  • Evident from the text: PM adjustment increased 95% posterior intervals dramatically and lost statistical significance, and the number of communities decreased as the ozone cutoff was lowered because some communities lacked sufficient data for model convergence or had no data at low ozone concentrations.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1289/ehp.10108

Cites

Statistics

Nothing to recompute.

Claims the paper makes

0 of 16 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    National and community-specific effect estimates of the short-term effects of ozone on mortality were robust to inclusion of PM10 or PM2.5 in time-series models.

    “National and community-specific effect estimates of the short-term effects of ozone on mortality were robust to inclusion of PM10 or PM2.5 in time-series models.”
    • Also stated in: § Inclusion of PM as a covariate in time-series models · plain
      “The national relative rate and individual community relative rate estimates are robust to inclusion of either PM metric as a covariate.”
    • Also stated in: § Discussion · boosted
      “This work also provides evidence that the relationship between short-term exposure to ozone and mortality is not confounded by adjustment for PM10 or PM2.5, at any strata of ozone or PM concentrations.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The subset of days with both ozone and PM data is sufficient and not systematically biased for estimating the ozone-mortality association.
    • Rests on an assumption: The 62 communities with sufficient PM2.5 data are representative of the 91 communities with PM2.5 data and of U.S. urban communities.
  • Findingown data

    The robustness of ozone-mortality effect estimates to PM adjustment remains even at low ozone levels (<10 ppb) using a subset approach.

    “The robustness remains even at low ozone levels (< 10 ppb) using a subset approach.”
    • Also stated in: § Subset approach with time-series models · hedged
      “Results indicate that community-specific and national effect estimates are robust to adjustment by PM10, for various values of s, including when only days with ozone levels < 10 ppb are considered.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The subset of days with both ozone and PM data is sufficient and not systematically biased for estimating the ozone-mortality association.
    • The authors state this as an assumption: If a threshold exists, effect estimates for ozone and mortality would be null for s values below the threshold.
  • Findingown data

    In the 91 communities with PM2.5 data, only 9.2% of study days had data on both ozone and PM2.5.

    “In the 91 communities with PM2.5 data, only 9.2% of days in the study period had data on ozone and PM2.5.”
    • Also stated in: § Frequency of data · plain
      “Only 9.2% of study days had both PM2.5 and ozone data on average across the 91 communities that measured PM2.5.”
    Quote in the paperSubstantial claim
  • Findingown data

    Ozone and PM10 were not highly correlated within strata of ozone concentrations; Pearson correlation coefficients ranged from <0.00 to 0.22.

    “Figure 1 indicates a lack of correlation between ozone and PM10 or PM2.5 levels at different strata of ozone concentrations.”
    • Also stated in: § Correlation between PM and ozone, stratified by pollution levels, season, and region · plain
      “The Pearson correlation coefficients ranged from < 0.00 to 0.22 for correlation within various strata of PM10”
    • Also stated in: § Discussion · plain
      “The Pearson correlation coefficients ranged from < 0.00 to 0.22 for correlation with PM10”
    Quote in the paperSubstantial claim
  • Findingown data

    PM2.5 data were available for 11.1% of study days on average across the 91 communities that measured PM2.5.

    “For PM2.5, data were available for 11.1% of the study days on average across the 91 communities that measured PM2.5.”
    Quote in the paperSubstantial claim
  • The quote only partly states this claim. 2
  • May combine two claims. 11
  • Classified as probably not a substantial claim. 9

  • The quote only partly states this claim.

    Findingown data

    Sensitivity analyses using different ozone metrics (daily 8-hr maximum and daily 1-hr maximum) and lag structures produced findings similar to the original same-day 24-hr ozone analysis.

    “These findings are similar to those in the original analysis based on same-day data using 24-hr ozone levels.”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Findingown data

    In strata of PM2.5 concentrations, the reported correlation coefficient between ozone and PM ranged from 0.03 to 0.10.

    “With PM2.5 strata of < 5, 5–10, 10–15, 15–20, 20–25, and > 25 μg/m3, the correlation coefficient between ozone and PM10 in each strata ranged from 0.03 to 0.10.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    On a national basis, no season demonstrated strong correlations between either PM metric and ozone levels; highest correlations were observed in spring and fall, and lowest in winter.

    “On a national basis, no season demonstrated strong correlations between either PM metric and ozone levels. Highest correlations were observed in spring and fall, and lowest in winter for both PM metrics, for the national data.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Using seasonally detrended data, correlation coefficients between PM2.5 and ozone within strata ranged from 0.02 to 0.21, and between PM10 and ozone ranged from 0.05 to 0.19.

    “The correlation coefficients between detrended PM2.5 and ozone within each strata ranged from 0.02 to 0.21, whereas the correlation coefficients between detrended PM10 and ozone ranged from 0.05 to 0.19.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Correlations between ozone and particles varied by region, with the highest correlations in the Industrial Midwest and Northeast for PM10 and in the Industrial Midwest and Southeast for PM2.5.

    “The correlations between ozone and particles varied by region, with the highest correlations in the Industrial Midwest and Northeast regions for PM10 and in the Industrial Midwest and Southeast for PM2.5.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    For days with low ozone concentrations (daily ozone <20 ppb), correlations were virtually zero for PM10 and negative for PM2.5.

    “For days with low ozone concentrations (daily ozone < 20 ppb), the correlations were virtually zero for PM10 and negative for PM2.5.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning

    The size of the 95% posterior intervals increases dramatically with PM adjustment, and statistical significance is lost.

    “These results demonstrate that although the overall national effect estimates are robust to PM adjustment, the size of the 95% posterior intervals increases dramatically with such analysis, and statistical significance is lost.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Ozone and PM2.5 were not highly correlated within strata of ozone concentrations; Pearson correlation coefficients ranged from −0.17 to 0.25.

    “The Pearson correlation coefficients ranged from < 0.00 to 0.22 for correlation within various strata of PM10 and −0.17 to 0.25 for correlation within various strata of PM2.5”
    • Also stated in: § Discussion · plain
      “The Pearson correlation coefficients ranged from < 0.00 to 0.22 for correlation with PM10 and −0.17 to 0.25 for correlation with PM2.5”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    PM is associated with increased risk of mortality for short-term exposure.

    “PM is also associated with increased risk of mortality for short-term exposure (Katsouyanni et al. 1997; Samet et al. 2000a) and could co-vary with ozone levels due to parallel sources of particles and ozone precursors (e.g., transportation).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Katsouyanni et al. 1997 source

      Not checked in the free run.

    • cites Samet et al. 2000a

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    A lag of 1 day for PM variables provided the most significant and largest effect in previous studies.

    “A lag of 1 day was used for PM variables because this lag structure provided the most significant and largest effect in previous studies (Peng et al. 2005).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Peng et al. 2005 source

      Not checked in the free run.

  • May combine two claims.

    Interpretationown data

    Results provide evidence that neither PM10 nor PM2.5 is a likely confounder of observed ozone and mortality relationships.

    “Results provide evidence that neither PM10 nor PM2.5 is a likely confounder of observed ozone and mortality relationships.”
    • Also stated in: § Discussion · hedged
      “In summary, we find that PM is unlikely to confound the ozone and mortality relationship.”
    • Also stated in: § Discussion · hedged
      “collectively our results support the hypothesis that PM does not confound the observed ozone and mortality associations.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The subset of days with both ozone and PM data is sufficient and not systematically biased for estimating the ozone-mortality association.
    • Rests on an assumption: The 62 communities with sufficient PM2.5 data are representative of the 91 communities with PM2.5 data and of U.S. urban communities.
    • The authors state this as an assumption: If a threshold exists, effect estimates for ozone and mortality would be null for s values below the threshold.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    A study of seven Pennsylvania and New Jersey counties from May 1992 to September 1995 found that peak ozone and mortality effect estimates were robust to adjustment by PM2.5 and PM10.

    “A study of seven Pennsylvania and New Jersey counties from May 1992 to September 1995 found that peak ozone and mortality effect estimates were robust to adjustment by PM2.5 and PM10 (Lipfert et al. 2000).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Lipfert et al. 2000 source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Effect estimates for PM2.5 and hospital admissions show temporal and spatial patterns, with the largest effects in the Northeast in summer.

    “Effect estimates for PM2.5 and hospital admissions show temporal and spatial patterns (Domenici et al. 2006), with the largest effects in the Northeast in summer (Peng et al. 2005), which could be the result of differential toxicity of particles.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Domenici et al. 2006

      Not checked in the free run.

    • cites Peng et al. 2005 source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Previous work on this issue has focused primarily on PM10, although strong mortality associations have been demonstrated for the fine fraction.

    “Previous work on this issue has focused primarily on PM10, although strong mortality associations have been demonstrated for the fine fraction (Kan et al. 2007; Ostro et al. 2006; Schwartz et al. 1996).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Kan et al. 2007 source

      Not checked in the free run.

    • cites Ostro et al. 2006 source

      Not checked in the free run.

    • cites Schwartz et al. 1996 source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Several previous time-series analyses suggest lack of evidence for confounding by PM10 or total suspended particles (TSP) including the United States and Europe, as well as single-city studies in Latin America and Korea.

    “Several previous time-series analyses suggest lack of evidence for confounding by PM10 or total suspended particles (TSP) including the United States and Europe, as well as single-city studies in Latin America and Korea (see discussion in U.S. EPA 2006).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites U.S. EPA 2006

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    In a meta-analysis, Levy et al. (2005) found that effect estimates for ozone and mortality were slightly higher in communities with higher annual or summer correlations between PM2.5 and ozone levels; however, the relationship was not statistically significant.

    “In a meta-analysis of previously conducted time-series studies of ozone and mortality, Levy et al. (2005) found that effect estimates for ozone and mortality were slightly higher in communities with higher annual or summer correlations between PM2.5 and ozone levels; however, the relationship was not statistically significant.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Levy et al. 2005 source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Previous analysis showed that the average of the same and previous day's ozone levels has higher effect estimates for ozone's relationship with mortality than other days.

    “Previous analysis showed that these days of exposure have higher effect estimates for ozone’s relationship with mortality than other days (Bell et al. 2004).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Bell et al. 2004 source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    Adjustment by PM2.5 was not applied for the subset approach due to sample size constraints.

    “Adjustment by PM2.5 was not applied for the subset approach due to sample size constraints.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    For comparability, the same data set was used for models with and without adjustment for PM.

    “For comparability, the same data set was used for models with and without adjustment for PM.”
    Quote in the paperSubstantial claim
19 of 19

  • The paper relies on this without support.

    The subset of days with both ozone and PM data is sufficient and not systematically biased for estimating the ozone-mortality association.

    “Only 25.0% of study days had data on both ozone and PM10, on average across the 93 communities with PM10 data. Only 9.2% of study days had both PM2.5 and ozone data on average across the 91 communities that measured PM2.5.”
    “Although data limitations restrict analysis of PM and ozone simultaneously, collectively our results support the hypothesis that PM does not confound the observed ozone and mortality associations.”

  • The paper relies on this without support.

    The 62 communities with sufficient PM2.5 data are representative of the 91 communities with PM2.5 data and of U.S. urban communities.

    “Although 91 communities had data for PM2.5, only 62 had sufficient data for both ozone and PM2.5 for the community-specific time-series analysis.”
    “Our results further show that the national effect estimate and community-specific relative rates of the short-term effects of ozone on mortality are robust to adjustment for PM2.5, based on the 62 U.S. communities with PM2.5 data available.”

  • The authors state this as an assumption.

    If a threshold exists, effect estimates for ozone and mortality would be null for s values below the threshold.

    “If a threshold value exists, effect estimates for ozone and mortality would be expected to be null for s values below the threshold.”
3 of 3

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

The citation lists haven’t been fetched yet

References 19 · Cited by 83

Research works printed but not linked: 6

  1. Anderson, 2004, Meta-Analysis of Time-Series Studies and Panel Studies of Particulate Matter (PM) and Ozone (O3)
  2. Dominici, 2005, The National Morbidity, Mortality, and Air Pollution Study. Part IV: Hierarchical Bivariate Time-Series Models—A Combined Analysis of PM10 Effects on Hospitalization and Mortality
  3. National Climatic Data Center (NCDC), Surface Airways. (n.d.) CD database
  4. Samet, 2000, The National Morbidity, Mortality, and Air Pollution Study
  5. U.S. EPA, 2006, Air Quality Criteria for Ozone and Related Photochemical Oxidants
  6. U.S. EPA, 2007, Air Quality System (AQS) Database

Authors
Bell, M. L., Kim, J. Y., & Dominici, F.
Affiliations
  • Yale University (US)
  • Environmental Protection Agency (US)
  • Johns Hopkins University (US)
Corresponding author
Bell, Michelle L.
ORCID
Venue
Environmental Health Perspectives
Publisher
American Chemical Society (ACS)
Published
2007-08-02
Volume
115
Issue
11
Pages
1591-1595
ISSN
  • 0091-6765 (ISSN-L)
  • 1552-9924
DOI
10.1289/ehp.10108
PMC
PMC2072830
OpenAlex
W2120948128
Semantic Scholar
d167403c0af2fe91dc77c3aee97c1e35a593a167
Cited by
83
References
19
Citations by year
2026: 1 · 2025: 2 · 2024: 2 · 2023: 4 · 2022: 5 · 2021: 4 · 2020: 4 · 2019: 2 · 2018: 6 · 2017: 3 · 2016: 1 · 2015: 5 · 2014: 6 · 2013: 11 · 2012: 6
FWCI
4.55
Type
journal article
Language
en
Open access
Licence
Locations
  • Environmental Health Perspectives (publishedVersion, public-domain)
  • PubMed Central (publishedVersion, public-domain)
  • PubMed (publishedVersion)
Funders
  • U.S. Environmental Protection Agency (RD-83241701)
  • Johns Hopkins University
  • National Center For Environmental Assessment
  • National Institute of Environmental Health Sciences (R01ES015028)
Topics
  • Physical Sciences › Environmental Science › Health, Toxicology and Mutagenesis › Air Quality and Health Impacts
  • Physical Sciences › Earth and Planetary Sciences › Atmospheric Science › Atmospheric chemistry and aerosols
  • Physical Sciences › Environmental Science › Health, Toxicology and Mutagenesis › Indoor Air Quality and Microbial Exposure
Keywords
ozone, particulate matter, PM10, PM2.5, confounding, short-term exposure, mortality, air pollution, urban communities, time series analysis, subgroup analysis
MeSH terms
Air Pollution, Humans, Mortality, Ozone, Time Factors, United States, Urban Population, Confounding Factors, Epidemiologic, Particulate Matter

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-10.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-10.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review