ASSESSING OPTIMAL INTERVENTION TARGETS FOR RESPIRATORY INFECTIONS IN STRUCTURED POPULATIONS

Thesis (Ph.D.)--University of Washington, 2019Background: Respiratory Viral Infections (RVI) are one of the most common health conditions globally, and are an enormous burden to health systems and society in terms of direct medical expenses and indirect productivity losses. Despite progress made in the 20th century with the introduction of antibiotics, vaccines, and antivirals there are no specific interventions for most respiratory infections of viral origin. There remains a need for new rationally designed therapeutics, and non-pharmaceutical interventions to prevent transmission. The objective of this project is improve current public health decision tools through an improved understanding of respiratory virus transmission from key ‘driver’ subgroups and their interactions with other subgroups in a study population. Methods: This research utilizes clinical epidemiology, infectious disease modeling, and extensive sensitivity analysis, to describe and evaluate current health decision-making processes related to two RVI. In the first chapter we infer exposure routes of Human parainfluenza virus-3 (HPIV-3) infection using a case-control study at a cancer-specific hospital. We examine locations and treatment exposures among patients that are associated with illness and make recommendations for the current intervention protocol based on the results. In the last two chapters we deal with previously quantified contact patterns. Using these contact patterns we evaluate the optimal vaccination intervention using cost-effectiveness analysis to compare the status quo influenza vaccination program to seven alternative strategies targeted at school age subgroups Preschool (2-4 years old), Primary school (5-11 years old), and Secondary school (12-16 years old). In the final chapter we deliberately incorporate social contact uncertainty into the same model as chapter 2 by using contact surveys from other countries and examine the variation in our cost-effectiveness results. Results: In our case-control study, 50 case subjects with medical appointments during their exposure period were compared to 106 controls using matched conditional logistic regression. Although contact precautions were enforced for immunosuppressed patients, these patients were 9.59 times as likely to be infected. Contact with several healthcare workers, especially nonclinical staff (e.g. social workers) was also associated with increased infection odds. In chapter 2, the incremental analysis demonstrated vaccinating 5-11 year olds (Primary School) was the most cost-efficient strategy. Although all seven strategies had a 100% probability of being cost-effective at the current National Health Service (NHS) threshold of £20,000 per 1 QALY gained, vaccination of 5-11 year olds provided substantial indirect protection to other age groups. Additionally strategies which contained this age group were less sensitive to changes in total achieved coverage. The structural uncertainty analysis in chapter 3 revealed Bayesian model fits using the substituted prior information on contact patterns did not converge on the same posterior. Substitution of a counterfactual contact structure in lieu of country-specific structure decreased the probability a strategy was cost-effective by propagating uncertainty through the model. We also determined that when a model’s contact structure is substituted the relative importance of each age group to population-level disease transmission varies substantially, rendering inconsistent cost-effectiveness results. Conclusion: In this research we evaluated interventions aimed at reducing two respiratory viral infections: seasonal influenza and HPIV-3. In the latter our results suggest asymptomatic shedding among hospital staff, especially non-clinical staff (e.g. social workers), and viral persistence in environmental surfaces as possible exposure routes of patient HPIV-3 infection. In chapter 2, we determined improvement could be made to the current LAIV pediatric vaccination strategy of England and Wales by eliminating LAIV vaccination of 2-4 year olds and focusing on school-based delivery of LAIV to two key age groups, Primary and Secondary school-age children. However the conclusions of chapter 2, and all policy interventions based on simulating health outcomes and economic costs, are sensitive to the scientific judgements around social contact parameters. From chapter 3 we conclude there is added value in using country-specific contact data, and advocate that in the absence of country-specific contact patterns, structural uncertainty analysis should be undertaken when quantifying the effect of strategies that involve herd immunity