Incorporating endogenous human behavior in models of COVID-19 transmission: A systematic scoping review.

Background: During the COVID-19 pandemic there was a plethora of dynamical forecasting models created, but their ability to effectively describe future trajectories of disease was mixed. A major challenge in evaluating future case trends was forecasting the behavior of individuals. When behavior was incorporated into models, it was primarily incorporated exogenously (e.g., fitting to cellphone mobility data). Fewer models incorporated behavior endogenously (e.g., dynamically changing a model parameter throughout the simulation). Methods: This review aimed to qualitatively characterize models that included an adaptive (endogenous) behavioral element in the context of COVID-19 transmission. We categorized studies into three approaches: 1) feedback loops, 2) game theory/utility theory, and 3) information/opinion spread. Findings: Of the 92 included studies, 72% employed a feedback loop, 27% used game/utility theory, and 9% used a model if information/opinion spread. Among all studies, 89% used a compartmental model alone or in combination with other model types. Similarly, 15% used a network model, 11% used an agent-based model, 7% used a system dynamics model, and 1% used a Markov chain model. Descriptors of behavior change included mask-wearing, social distancing, vaccination, and others. Sixty-eight percent of studies calibrated their model to observed data and 25% compared simulated forecasts to observed data. Forty-one percent of studies compared versions of their model with and without endogenous behavior. Models with endogenous behavior tended to show a smaller and delayed initial peak with subsequent periodic waves. Interpretation: While many COVID-19 models incorporated behavior exogenously, these approaches may fail to capture future adaptations in human behavior, resulting in under- or overestimates of disease burden. By incorporating behavior endogenously, the next generation of infectious disease models could more effectively predict outcomes so that decision makers can better prepare for and respond to epidemics. Funding: This study was funded in-part by Centers for Disease Control and Prevention (CDC) MInD-Healthcare Program (1U01CK000536), the National Science Foundation (NSF) Modeling Dynamic Disease-Behavior Feedbacks for Improved Epidemic Prediction and Response grant (2229996), and the NSF PIPP Phase I: Evaluating the Effectiveness of Messaging and Modeling during Pandemics grant (2200256).

Modeling the potential health impacts of delayed strain selection on influenza hospitalization and mortality with mRNA vaccines.

Background: Influenza viruses are constantly evolving through antigenic drift, which makes vaccines potentially ill-matched to circulating strains due to the time between strain selection and distribution. mRNA technology could improve vaccine effectiveness (VE) by reducing this time. Significant private and public investments would be required to accommodate accelerated vaccine development and approval. Hence, it is important to understand the potential impact of mRNA technology on influenza hospitalizations and mortality. Methods: We developed an age-stratified dynamic model of influenza transmission to evaluate the potential impact of increased VE (increased protection against either infection or only hospitalization) on hospitalizations and mortality in the United States. We assume that mRNA technology allows for delaying the time to strain choice, which might increase efficacy, but it does not reduce the time needed for distribution and administration, which might reduce availability. To assess this tradeoff, we evaluated two scenarios where strain choice was delayed until late summer resulting in a more effective vaccine available to (1) all age groups by October, or (2) adults 65 years and older starting in August. Results: If not available until October, the vaccine would need a minimum of 95% effectiveness against infection to see a decrease in hospitalizations and deaths in all age groups. When delayed until November, even a 100% effective vaccine had no significant impact. For the elderly, the minimum required VE (against infection) was 50% to reduce hospitalizations and deaths. Moreover, a vaccine with 80% VE against infection available in August for the 65 + age group was better than a 95% effective vaccine available in October for all ages. Conclusions: As the majority of influenza-associated hospitalizations and deaths are in adults 65 years and older, a combination policy targeting higher VE and coverage for this age group in the short term would be the most efficacious.

COVID-19 and Outpatient Antibiotic Prescriptions in the United States: A County-Level Analysis.

Background: Declines in outpatient antibiotic prescribing were reported during the beginning of the coronavirus disease 2019 (COVID-19) pandemic in the United States; however, the overall impact of COVID-19 cases on antibiotic prescribing remains unclear. Methods: This was an ecological study using random-effects panel regression of monthly reported COVID-19 county case and antibiotic prescription data, controlling for seasonality, urbanicity, health care access, nonpharmaceutical interventions (NPIs), and sociodemographic factors. Results: Antibiotic prescribing fell 26.8% in 2020 compared with prior years. Each 1% increase in county-level monthly COVID-19 cases was associated with a 0.009% (95% CI, 0.007% to 0.012%; P < .01) increase in prescriptions per 100 000 population dispensed to all ages and a 0.012% (95% CI, -0.017% to -0.008%; P < .01) decrease in prescriptions per 100 000 children. Counties with schools open for in-person instruction were associated with a 0.044% (95% CI, 0.024% to 0.065%; P < .01) increase in prescriptions per 100 000 children compared with counties that closed schools. Internal movement restrictions and requiring facemasks were also associated with lower prescribing among children. Conclusions: The positive association of COVID-19 cases with prescribing for all ages and the negative association for children indicate that increases in prescribing occurred primarily among adults. The rarity of bacterial coinfection in COVID-19 patients suggests that a fraction of these prescriptions may have been inappropriate. Facemasks and school closures were correlated with reductions in prescribing among children, possibly due to the prevention of other upper respiratory infections. The strongest predictors of prescribing were prior years' prescribing trends, suggesting the possibility that behavioral norms are an important driver of prescribing practices.

Global antimicrobial-resistance drivers: an ecological country-level study at the human-animal interface.

BACKGROUND: Antimicrobial resistance (AMR) is a pressing, holistic, and multisectoral challenge facing contemporary global health. In this study we assessed the associations between socioeconomic, anthropogenic, and environmental indicators and country-level rates of AMR in humans and food-producing animals. METHODS: In this modelling study, we obtained data on Carbapenem-resistant Acinetobacter baumanii and Pseudomonas aeruginosa, third generation cephalosporins-resistant Escherichia coli and Klebsiella pneumoniae, oxacillin-resistant Staphylococcus aureus and vancomycin-resistant Enterococcus faecium AMR in humans and food-producing animals from publicly available sources, including WHO, World Bank, and Center for Disease Dynamics Economics and Policy. AMR in food-producing animals presented a combined prevalence of AMR exposure in cattle, pigs, and chickens. We used multivariable ß regression models to determine the adjusted association between human and food-producing animal AMR rates and an array of ecological country-level indicators. Human AMR rates were classified according to the WHO priority pathogens list and antibiotic-bacterium pairs. FINDINGS: Significant associations were identified between animal antimicrobial consumption and AMR in food-producing animals (OR 1·05 [95% CI 1·01-1·10]; p=0·013), and between human antimicrobial consumption and AMR specifically in WHO critical priority (1·06 [1·00-1·12]; p=0·035) and high priority (1·22 [1·09-1·37]; p<0·0001) pathogens. Bidirectional associations were also found: animal antibiotic consumption was positively linked with resistance in critical priority human pathogens (1·07 [1·01-1·13]; p=0·020) and human antibiotic consumption was positively linked with animal AMR (1·05 [1·01-1·09]; p=0·010). Carbapenem-resistant Acinetobacter baumanii, third generation cephalosporins-resistant Escherichia coli, and oxacillin-resistant Staphylococcus aureus all had significant associations with animal antibiotic consumption. Analyses also suggested significant roles of socioeconomics, including governance on AMR rates in humans and animals. INTERPRETATION: Reduced rates of antibiotic consumption alone will not be sufficient to combat the rising worldwide prevalence of AMR. Control methods should focus on poverty reduction and aim to prevent AMR transmission across different One Health domains while accounting for domain-specific risk factors. The levelling up of livestock surveillance systems to better match those reporting on human AMR, and, strengthening all surveillance efforts, particularly in low-income and middle-income countries, are pressing priorities. FUNDING: None.

Modeling of malaria vaccine effectiveness on disease burden and drug resistance in 42 African countries.

BACKGROUND: The emergence of antimalarial drug resistance poses a major threat to effective malaria treatment and control. This study aims to inform policymakers and vaccine developers on the potential of an effective malaria vaccine in reducing drug-resistant infections. METHODS: A compartmental model estimating cases, drug-resistant cases, and deaths averted from 2021 to 2030 with a vaccine against Plasmodium falciparum infection administered yearly to 1-year-olds in 42 African countries. Three vaccine efficacy (VE) scenarios and one scenario of rapidly increasing drug resistance are modeled. RESULTS: When VE is constant at 40% for 4 years and then drops to 0%, 235.7 (Uncertainty Interval [UI] 187.8-305.9) cases per 1000 children, 0.6 (UI 0.4-1.0) resistant cases per 1000, and 0.6 (UI 0.5-0.9) deaths per 1000 are averted. When VE begins at 80% and drops 20 percentage points each year, 313.9 (UI 249.8-406.6) cases per 1000, 0.9 (UI 0.6-1.3) resistant cases per 1000, and 0.9 (UI 0.6-1.2) deaths per 1000 are averted. When VE remains 40% for 10 years, 384.7 (UI 311.7-496.5) cases per 1000, 1.0 (0.7-1.6) resistant cases per 1000, and 1.1 (UI 0.8-1.5) deaths per 1000 are averted. Assuming an effective vaccine and an increase in current levels of drug resistance to 80% by 2030, 10.4 (UI 7.3-15.8) resistant cases per 1000 children are averted. CONCLUSIONS: Widespread deployment of a malaria vaccine could substantially reduce health burden in Africa. Maintaining VE longer may be more impactful than a higher initial VE that falls rapidly.

Malaria can become resistant to the drugs used to treat it, posing a major threat to malaria treatment and control. An effective vaccine has the potential to reduce both resistant infections and antimalarial drug use. However, how successfully a vaccine can protect against infection (vaccine efficacy) and the impact of increasing drug resistance remain unclear. Using a mathematical model, we estimate the impact of malaria vaccination in 42 African countries over a 10-year period in multiple scenarios with differing vaccine efficacy and drug resistance. Our model suggests that a moderately effective vaccine with sustained protection over a long period could avert more resistant infections and deaths than a vaccine that is highly protective initially but lowers in efficacy over time. Nevertheless, implementation of an effective malaria vaccine should be accelerated to mitigate the health and economic burden of drug resistance.

Investigating the effects of absolute humidity and movement on COVID-19 seasonality in the United States.

Mounting evidence suggests the primary mode of SARS-CoV-2 transmission is aerosolized transmission from close contact with infected individuals. While transmission is a direct result of human encounters, falling humidity may enhance aerosolized transmission risks similar to other respiratory viruses (e.g., influenza). Using Google COVID-19 Community Mobility Reports, we assessed the relative effects of absolute humidity and changes in individual movement patterns on daily cases while accounting for regional differences in climatological regimes. Our results indicate that increasing humidity was associated with declining cases in the spring and summer of 2020, while decreasing humidity and increase in residential mobility during winter months likely caused increases in COVID-19 cases. The effects of humidity were generally greater in regions with lower humidity levels. Given the possibility that COVID-19 will be endemic, understanding the behavioral and environmental drivers of COVID-19 seasonality in the United States will be paramount as policymakers, healthcare systems, and researchers forecast and plan accordingly.

Machine learning and artificial intelligence: applications in healthcare epidemiology.

Artificial intelligence (AI) refers to the performance of tasks by machines ordinarily associated with human intelligence. Machine learning (ML) is a subtype of AI; it refers to the ability of computers to draw conclusions (ie, learn) from data without being directly programmed. ML builds from traditional statistical methods and has drawn significant interest in healthcare epidemiology due to its potential for improving disease prediction and patient care. This review provides an overview of ML in healthcare epidemiology and practical examples of ML tools used to support healthcare decision making at 4 stages of hospital-based care: triage, diagnosis, treatment, and discharge. Examples include model-building efforts to assist emergency department triage, predicting time before septic shock onset, detecting community-acquired pneumonia, and classifying COVID-19 disposition risk level. Increasing availability and quality of electronic health record (EHR) data as well as computing power provides opportunities for ML to increase patient safety, improve the efficiency of clinical management, and reduce healthcare costs.