This thesis presents the application of three mathematical models to problems linking environmental exposures to human health. The models differ in spatial and temporal analysis scale. The premise underlying this work is that reliable models follow from careful matching of model scale to the specific research question.
Chapter 1 models bacterial competition at a cellular scale, to study the factors that may result in environmental antimicrobial resistance. A simple analytical solution for the antibiotic minimum selection concentration (MSC) is developed. The MSC is the lowest environmental antibiotic concentration at which a resistant bacterial strain will outcompete a sensitive strain. The solution is formulated as the ratio between the MSC and the minimum inhibitory concentration (MIC), which is a widely available laboratory measurement of the antibiotic concentration at which the growth of a sensitive strain is inhibited. Model equations were fitted to published experimental growth rate competition results. The model fit varied among nine compound-taxa combinations examined, but predicted the experimentally observed MSC/MIC ratio well (R2 ≥ 0.95). Sensitivity analysis indicated that the MSC was sensitive to the shape of the antibiotic versus growth dose–response for the sensitive strain and to the fitness difference between strains. Model findings suggest a benefit of future experimental studies characterizing bacterial competition at low antibiotic concentrations. Employing the model in combination with empirical antibiotic growth curve data, it may be possible to predict environmental antibiotic concentrations at which resistant strains will be selected for. This could be incorporated into risk assessment models, to identify high risk environments for dissemination of antibiotic resistance.
Chapter 2 describes a quantitative model of the relative importance of direct skin-to-skin contact versus indirect transfer via environmental textiles and surfaces for hospital pathogens. The model describes the rate of environmental transfer of pathogenic microbes between patients in a hospital setting. However, the model does not consider the likelihood of infection. The model was applied to transmission of pathogens between patients residing in separate hospital rooms, via a health-care worker. Simulations were performed to examine the separate contribution of skin, textiles, and nonporous surfaces to the total pathogen number transmitted. The role of elimination (organism death) was considered by comparing literature elimination rates for six pathogens: Acinetobacter baumannii, Staphylococcus aureus, Streptococcus pneumoniae, Bordetella pertussis, sudden acute respiratory syndrome coronavirus (SARS-CoV), and influenza A. Based on model results, all pathogens except influenza A exhibit a high rate of transmission in the model scenario, suggesting that transmission via health-care workers is a valid concern. With the exception of influenza A, there was overlap in literature elimination rates among the pathogens, resulting in similarly high predicted transmission. For all pathogens except SARS-CoV the relative importance for pathogen transmission was nonporous surfaces > textiles > skin, indicating the importance of environmental surfaces as a potential pathway for disease transmission. For SARS-CoV, the order was nonporous surfaces > skin > textiles, due to literature indicating low survival on textiles and porous surfaces. These results, combined with limited data on elimination, suggest a need to perform disease-specific studies on how elimination systematically differs between skin and surfaces. This model application at the scale of individual humans indicates that environmental surfaces are likely important for pathogen transmission in health care settings.
Chapter 3 describes multivariate and geostatistical modeling employed to perform a combined assessment of multiple stressors at a regional scale. The study evaluated a metric of environmental health hazard developed by the California Environmental Protection Agency. The metric, CalEnviroScreen, combines 19 indicators of environmental impact and socioeconomic stress, and is intended to be used to help allocate funding for greenhouse gas amelioration projects within the state of California. Principal component analysis was performed to obtain the predominant multivariate associations in the 19 indicators. The CalEnviroScreen metric was strongly associated with the first principal components, indicating that CalEnviroScreen effectively captures the prevailing gradients in hazard present in the underlying data. However, CalEnviroScreen was poorly associated with agricultural pesticide application, suggesting that hazard from agricultural chemical exposure may not be captured. The first principal components obtained from the environmental pollution measures and the socioeconomic stressor measures were both associated with the rate of hospital visits for several disease diagnoses with an environmental etiology. This suggests that the indicators employed for CalEnviroScreen are associated with the burden of disease. The association was stronger for socioeconomic stressors than for environmental pollutants. The results of this ecological health study suggest a hypothesis that, compared to environmental pollutant exposure, socioeconomic status more greatly impacts overall burden of disease.
