John Glasser

John Glasser studied biology at Princeton, population biology at Duke, and international health, biostatistics, and epidemiology at Harvard. After Epidemic Intelligence Service in the CDC’s Division of Reproductive Health, he returned to Harvard to study mathematical biology with the late Richard Levins. By comparing scenarios differing solely in phenomena of interest, realistic mathematical modeling is among the most reliable, if not only means of evaluating public health programs. Such evaluations can be prospective, involving hypothetical interventions being contemplated, or retrospective, involving alternatives to ongoing programs. Initially at the CDC and currently at Emory, he has assisted in designing, or evaluating and occasionally improving public health policy at home and abroad by modeling the transmission of multiple pathogens among human hosts, most causing vaccine-preventable respiratory illnesses. He and his colleagues helped to formulate the US response were Variola major reintroduced by terrorists. And they explained why earlier care-seeking for symptoms that might herald SARS, together with increasingly accurate diagnoses and effective isolation, had far more impact than quarantine. As mixing is the essence of meta-population modeling, they generalized Annette Nold’s (Math Biosci 1980; 52:227-40) function to include preferential contacts between parents and children as well as contemporaries and age-independent contacts among co-workers. They also developed a method for estimating age-specific rates of infection from cross-sectional serological surveys when passively acquired maternal antibodies decay and active immunity wanes, allowing re-infection with clinical consequences that depend on residual immunity. Assuming that vaccination is responsible for secular changes in the epidemiology of pertussis throughout the developed world – by reducing the exposures to infectious children that used to boost immunity – they deduced the optimal number and timing of revaccinations in Sweden. They also explored the impact of heterogeneity in factors affecting effective reproduction numbers, and hence our ability to prevent or control outbreaks, and applied their insights to spatial heterogeneity in vaccine coverage due to personal-belief exemptions, together with preferential mixing among like-minded people. And, perhaps most significantly, they promoted gradients of effective reproduction numbers (partial derivatives with respect to control parameters) as means of identifying optimal outbreak prevention or control measures. Recently, they devised vaccination strategies for accelerating elimination of measles, rubella and other vaccine-preventable diseases from China, and compared strategies for mitigating COVID-19 disease and SARS-CoV-2 transmission in the US. Ongoing projects include comparing respiratory disease burden and mitigation measure impact assessments with surveillance-based ones, identifying means by which novel SARS-CoV-2 variants most likely emerged, and relaxing unrealistic assumptions in their gradient calculations. Finally, they have become concerned about fitting transmission model parameters to disease surveillance, as this is tantamount to assuming that models and observations are both correct. They consider mechanistic models to be hypotheses, compare their predictions to accurate independent observations, and remedy the cause(s) of any disparities before exploring contrafactual scenarios. But modelers trained in quantitative disciplines, few of which include philosophy of science, evidently are unaware that while fitting is necessary for descriptive models, it precludes testing mechanistic ones. Their efforts to remedy this problem, which is ethical – insofar as few health policymakers can evaluate mathematical models – as well as epistemological, include an essay (Math Biosci 2025; 383:109419), mini-symposium at this year’s Society for Mathematical Biology meeting, and planned workshop.