Management of Environmental Risks from Air Pollution Exposure on Cardiovascular and Respiratory Health Using Hybrid Predictive and Optimization Framework

Abstract

Air pollution is a serious health problem in the world especially to those who
already have cardiorespiratory illnesses. Although the association between
exposure to pollution and health exacerbation is well-established, the current
interventions in the area of public health are not customized by taking into account
personal vulnerability, geographic position, and daily routine activities. Existing
studies are mainly concentrated on predictive modeling based on machine
learning or macro-level resource optimization, but do not combine these two
measures into practical and personalized guidance. This work fills this critical gap
by hypothesizing and confirming the new hybrid machine learning-optimization
model to produce individual alerts on air pollution exposures. An artificial group
of 100 patients with asthma, COPD and ischemic heart disease is modeled in 90
days. The first stage entails the training of a logistic regression model to estimate
short-term exacerbation risk (with an AUC of 0.979 and an accuracy of 0.961).
This risk score is then inputted into a second-stage mixed-integer linear
programming resulting in the production of optimal daily action plans. The
findings indicate that the framework effectively orders viable interventions in a
successful completion of an average personal exposure reduction of 39.4 percent
of all case studies without violating individual involvement restrictions and
interests, making it a viable paradigm of precision environmental health.