Cases of improvement to public health systems using mathematical modeling
Davila Payan, Carlo Stefan
MetadataShow full item record
This work builds on the use of several Mathematical Modeling tools to develop approaches that address relevant, real and previously unanswered questions related to the improvement of Public Health Systems, in three particular instances. First, this thesis analyzes the variation in state-level vaccination coverage during the emergency response to the 2009 H1N1 pandemic influenza outbreak in the United States. The analysis considers the overall adults population and two priority sub-populations: children and high-risk adults. We focus on quantifying the association between vaccination coverage and the supply chain and distribution system decisions, during the vaccine shortage period, while controlling for other commonly recognized factors such as previous vaccinations, socio-economic characteristics, health seeking behavior and health infrastructure. The variables analyzed are generally correlated, and the problem has a limited sample size with a much larger number of independent variables. The findings of this research have been published in Vaccine and presented to the Centers for Disease Control and Prevention. Second, the research approaches the problem of estimating childhood obesity prevalence in small geographic areas in the U. S. Obesity is recognized as one of the major health problems in the country, and attending this condition in children is of major importance to deal with the sources of the overall problem. The ability to target interventions to the most affected children populations is necessary to achieve cost effective solutions. But local accurate obesity data is hard to obtain and missing for most of the small areas in the country. The research focuses on estimating prevalence of obesity and overweight status in children in small geographical areas in the absence of surveillance and detailed sampling. Our modeling approach is built in two stages. The first one uses a logistic regression model that links individual characteristics to high-BMI status, and generates samples of the empirical distribution of its coefficients though bootstrap re-sampling. The second uses simulation to generate virtual population samples of the small areas, which are then combined with the logistic model samples to estimate prevalence. Confidence intervals are built though re-sampling. A very important feature of our approach is that all of its inputs are from publicly available data, which gives availability for the replication of the methodology to any health stakeholder in the US. The model estimates were validated by using separate models for adults and children in a state with available data. Estimates obtained from our modeling approach were used by a large healthcare provider to geographically target interventions for pediatric obesity. Third, the thesis presents an introductory analysis of the possible effects of partial disruptions to critical supply chains due to absenteeism caused by a generalized flu-like illness in the US. For this analysis, we first construct a plausible national food supply chain for milk and then we simulate its disruption. To build the supply chain we used public information regarding production, consumption, and major milk processors and bottlers, and fitted it into a supply network though optimization. Then, to analyze the effects of flow disruptions of the supply chain, we built a simulation of the operation of the network and virtually generated absenteeism, mildly disrupting the supply chain flows by the proportional absences. We used information on potential absenteeism in work groups from an influenza simulator. Our initial analysis shows that absenteeism may create variations along the supply chain, similar to those described in the bullwhip effect analysis literature, even in the absence of supply shortages and without variations in pricing or demand, for which we find no prior reference in the literature.