William Dana Flanders

Background: Residual confounding is challenging to detect. Recently, we described a method for detecting confounding and justified it primarily for time-series studies. The method depends on an indicator with 2 key characteristics: (1) it is conditionally independent (given measured exposures and covariates) of the outcome, in the absence of confounding, misspecification, and measurement errors; and (2) like the exposure, it is associated with confounders, possibly unmeasured. We proposed using future exposure levels as the indicator to detect residual confounding. This choice seems natural for time-series studies because future exposure cannot have caused the event, yet they could be spuriously related to it. A related question addressed here is whether an analogous indicator can be used to identify residual confounding in a study based on spatial, rather than temporal, contrasts. Methods: Using directed acyclic graphs, we show that future air pollution levels may have the characteristics appropriate for an indicator of residual confounding in spatial studies of environmental exposures. We empirically evaluate performance for spatial studies using simulations. Results: In simulations based on a spatial study of ambient air pollution levels and birth weight in Atlanta, and using ambient air pollution 1 year after conception as the indicator, we were able to detect residual confounding. The discriminatory ability approached 100% for some factors intentionally omitted from the model, but was very weak for others. Conclusion: The simulations illustrate that an indicator based on future exposures can have excellent ability to detect residual confounding in spatial studies, although performance varied by situation.

Background: A difficult issue in observational studies is assessment of whether important confounders are omitted or misspecified. In this study, we present a method for assessing whether residual confounding is present. Our method depends on availability of an indicator with 2 key characteristics: first, it is conditionally independent (given measured exposures and covariates) of the outcome in the absence of confounding, misspecification, and measurement errors; second, it is associated with the exposure and, like the exposure, with any unmeasured confounders. Methods: We demonstrate the method using a time-series study of the effects of ozone on emergency department visits for asthma in Atlanta. We argue that future air pollution may have the characteristics appropriate for an indicator, in part because future ozone cannot have caused yesterday's health events. Using directed acyclic graphs and specific causal relationships, we show that one can identify residual confounding using an indicator with the stated characteristics. We use simulations to assess the discriminatory ability of future ozone as an indicator of residual confounding in the association of ozone with asthma-related emergency department visits. Parameter choices are informed by observed data for ozone, meteorologic factors, and asthma. Results: In simulations, we found that ozone concentrations 1 day after the emergency department visits had excellent discriminatory ability to detect residual confounding by some factors that were intentionally omitted from the model, but weaker ability for others. Although not the primary goal, the indicator can also signal other forms of modeling errors, including substantial measurement error, and does not distinguish between them. Conclusions: The simulations illustrate that the indicator based on future air pollution levels can have excellent discriminatory ability for residual confounding, although performance varied by situation. Application of the method should be evaluated by considering causal relationships for the intended application, and should be accompanied by other approaches, including evaluation of a priori knowledge.

Substantial evidence supports an association between ambient air pollution, especially particulate matter (PM*) and ozone (O3), and acute cardiovascular and respiratory morbidity. There is increasing interest in accountability research to evaluate whether actions taken to reduce air pollution will result in reduced morbidity. This study capitalized on a unique opportunity to evaluate the impact of a local, short-term intervention effort to reduce traffic in Atlanta during the 1996 Summer Olympic Games (July 19-August 4). Air pollutant concentrations both inside and outside of Atlanta were examined during the Olympic period and surrounding periods. Emergency department (ED) visits were examined to evaluate changes in usage patterns. ED visits for respiratory and cardiovascular conditions were examined in relation to the Olympic period using Poisson time-series analysis with adjustment for time trends and meteorologic conditions. O3 concentrations were approximately 30% lower during the Olympic Games compared with the four weeks before and after the Olympic Games (baseline periods); however, we observed similar reductions in O3 concentrations in several other cities in the Southeastern United States. We observed little or no evidence of reduced ED visits during the Olympic Games; the estimates were sensitive to choice of analytic model and to method of adjusting for temporal trends. The meteorologic conditions during the Olympic Games, along with the reductions in O3 observed in various cities not impacted by the Olympic Games, suggest that both meteorologic conditions-and reduced traffic may have played a role in the observed reduction in O3 concentration in Atlanta. Additionally, it is likely that this particular intervention strategy would not be sustainable as a pollution-reduction strategy. This study demonstrates some limitations of conducting retrospective accountability research.

Reverse causality, in which obesity-induced disease leads to both weight loss and higher mortality, may bias observed associations between body mass index (BMI) and mortality, but the magnitude of that bias is unknown. The authors examined the impact of reverse causality and the exclusion of various diseases on the observed age-specific mortality ratios for BMI by using a state space model and sensitivity analyses. They found that reverse causality may decrease the ratios and induce a J-shaped curve on a graph. The authors further found that the net effect of excluding various diseases becomes a balance of competing forces, some tending to increase observed mortality ratios, where as others, such as selection based on common effects, may decrease them. Instead of studying just the change in observed mortality ratios, which can be misleading, investigators need to consider causal relationships and evaluate the conceptual and theoretical impact of any analytic maneuver. Analyses should be balanced with sensitivity approaches as well as with alternative analytic approaches such as the use of structural models, G-estimation, simulations and ancillary data from animal studies.

As recognized for more than 2 decades, the way to define effects of an exposure may be unclear if the effects are conditional on occurrence of prior events. Since age-specific rates are inherently conditional on survival to the age for which rates are calculated, age-specific rate ratios may be misleading. We consider this problem in the context of a point exposure and an unmeasured risk factor that is independent of the exposure, together with potential outcome models and associated counterfactual effect definitions. The methods apply to a recurring exposure that "tracks" over time, as well as to more complicated situations (although additional issues may then arise). We identify and evaluate 2 seemingly-natural ways that the population effects of a point exposure might be defined. At least one definition of the population effects of a point exposure is identifiable, while another natural definition is not identifiable. We describe possible implications of these definitions for the distortion of time-specific rate ratios that can occur with passage of time. We discuss interpretation of effects for each definition, and how the definitions are related to selection bias as recently defined by Hernán et al (Epidemiology. 2004;15:615-625). We present implications for study design, and make several recommendations. Problems may be reduced or avoided by starting follow-up before onset of exposure and by using survival curves to compare exposed with unexposed.

The sufficient-component cause (SSC) model provides a useful, concise method to conceptualize and organize ideas concerning biologic effects and interactions of multiple factors. Another type of model, the potential outcome or counterfactual models, has also proved useful in epidemiology, providing insights into definitions of effects, exchangeability and confounding, and selection bias. Prior work has shown important links between these two types of models. Here, we show additional connections between the two types of models. We first review basic concepts for both the SSC model and the potential outcome models. We then indicate additional similarities between them, and derive a quantitative link. Our results show that the SSC model actually corresponds to a unique response pattern in a particular potential outcome model. Recognition of the links between the two models should allow greater insight into use of each model, and allow the strengths of both models to be used jointly.

The aim of this study was to analyze the association between Physical Activity (PA) and forearm bone mineral density (BMD) in healthy postmenopausal women. During 1984–1986, a population-based health survey (HUNT-1) was conducted in Nord-Trøndelag county, Norway. The second follow-up survey (HUNT-2) was conducted during 1995–1997. The subjects consist of all healthy postmenopausal women (N = 2,924). Higher intensity of recreational PA in HUNT-1 was associated with higher BMD ultradistally (slope = +0.0084, p = 0.0009). The summary score of duration, frequency and intensity of recreational PA in HUNT-1 was associated with lower risk of being below the 20th percentile ultradistally (OR = 0.90, p = 0.01, after adjustment for covariates). The trends for mean distal and ultradistal BMD and percent low BMD lacked statistical significance when we used the PA score that combined the recreational and occupational PA in HUNT-1 and HUNT-2. Conclusion: Higher intensity of prior recreational PA (HUNT-1) was associated with a protective effect on BMD measured in the forearm (HUNT-2).