Robust and Equitable Public Health Screening Strategies, with Application to Genetic and Infectious Diseases
Public health screening plays an important role in the overall healthcare system. As an example, consider newborn screening, a state-level initiative that screens newborns for life-threatening genetic disorders for which early treatment can substantially improve health outcomes. Another topical example is in the realm of infectious disease screening, e.g., screening for COVID-19. The common features of both public health screening problems include large testing populations and resource limitations that inhibit screening efforts. Cost is a major barrier to the inclusion of genetic disorders in newborn screening, and thus screening must be both highly accurate and efficient; and for COVID-19, limited testing kits, and other shortages, have been major barriers to screening efforts. Further, for both newborn screening and infectious disease screening, equity (reducing health disparities among different sub-populations) is an important consideration.
We study the testing process design for newborn screening for genetic diseases, considering cystic fibrosis as a model disorder. Our optimization-based models take into account disease-related parameters, subject risk factors, test characteristics, parameter uncertainty, and limited testing resources so as to design equitable, accurate, and robust screening processes that classify newborns as positive or negative for cystic fibrosis. Our models explicitly consider the trade-off between false-negatives, which lead to missed diagnoses, and the required testing resources; and the trade-off between the accuracy and equity of screening. We also study the testing process design for infectious disease screening, considering COVID-19 as a model disease. Our optimization-based models account for key subject risk factors that are important to consider, including the likelihood of being disease-positive, and the potential harm that could be averted through testing and the subsequent interventions. Our objectives include the minimization of harm (through detection and mitigation) or maximization of testing coverage.
These are complex problems. We develop novel mathematical models and characterize key structural properties of optimal solutions. This, in turn, allows the development of effective and efficient algorithms that exploit these structural properties. These algorithms are either polynomial- or pseudo-polynomial-time algorithms, and are able to solve realistic-sized problems efficiently. Our case studies on cystic fibrosis screening and COVID-19 screening, based on realistic data, underscore the value of the proposed optimization-based approaches for public health screening, compared to current practices. Our findings have important implications for public policy.