College campuses and COVID-19 mitigation: clinical and economic value (preprint)

BackgroundDecisions around US college and university operations will affect millions of students and faculty amidst the COVID-19 pandemic. We examined the clinical and economic value of different COVID-19 mitigation strategies on college campuses. MethodsWe used the Clinical and Economic Analysis of COVID-19 interventions (CEACOV) model, a dynamic microsimulation that tracks infections accrued by students and faculty, accounting for community transmissions. Outcomes include infections, $/infection-prevented, and $/quality-adjusted-life-year ($/QALY). Strategies included extensive social distancing (ESD), masks, and routine laboratory tests (RLT). We report results per 5,000 students (1,000 faculty) over one semester (105 days). ResultsMitigation strategies reduced COVID-19 cases among students (faculty) from 3,746 (164) with no mitigation to 493 (28) with ESD and masks, and further to 151 (25) adding RLTq3 among asymptomatic students and faculty. ESD with masks cost $168/infection-prevented ($49,200/QALY) compared to masks alone. Adding RLTq3 ($10/test) cost $8,300/infection-prevented ($2,804,600/QALY). If tests cost $1, RLTq3 led to a favorable cost of $275/infection-prevented ($52,200/QALY). No strategies without masks were cost-effective. ConclusionExtensive social distancing with mandatory mask-wearing could prevent 87% of COVID-19 cases on college campuses and be very cost-effective. Routine laboratory testing would prevent 96% of infections and require low cost tests to be economically attractive.

Management Strategies for People Experiencing Sheltered Homelessness during the COVID-19 Pandemic: Clinical Outcomes and Costs (preprint)

ABSTRACT Importance: Approximately 356,000 people stay in homeless shelters nightly in the US. These individuals are at high risk for COVID-19. Objective: To assess clinical outcomes, costs, and cost-effectiveness of strategies for COVID-19 prevention and management among sheltered homeless adults. Design: We developed a dynamic microsimulation model of COVID-19. We modeled sheltered homeless adults in Boston, Massachusetts, using cohort characteristics and costs from Boston Health Care for the Homeless Program. Disease progression, transmission, and clinical outcomes data were from published literature and national databases. We examined surging, growing, and slowing epidemics (effective reproduction numbers [Re] 2.6, 1.3, and 0.9). Costs were from a health care sector perspective; time horizon was 4 months. Setting & Participants: Simulated cohort of 2,258 adults residing in homeless shelters in Boston. Interventions: We assessed combinations of daily symptom screening with same-day polymerase chain reaction (PCR) testing of screen-positive individuals, universal PCR testing every 2 weeks, hospital-based COVID-19 care, alternate care sites [ACSs] for mild/moderate COVID-19 management, and moving people from shelters to temporary housing, compared to no intervention. Main Outcomes: Infections, hospital-days, costs, and cost-effectiveness. Results: Compared to no intervention, daily symptom screening with ACSs for those with pending tests or confirmed COVID-19 and mild/moderate disease leads to 37% fewer infections and 46% lower costs when Re=2.6, 75% fewer infections and 72% lower costs when Re=1.3, and 51% fewer infections and 51% lower costs when Re=0.9. Adding universal PCR testing every 2 weeks further decreases infections in all epidemic scenarios, with incremental cost per case prevented of $1,000 (Re=2.6), $27,000 (Re=1.3), and $71,000 (Re=0.9). In all scenarios, moving shelter residents to temporary housing with universal PCR testing every 2 weeks is most effective but substantially more costly than other options. Results are most sensitive to the cost and sensitivity of PCR testing and the efficacy of ACSs in preventing transmission. Conclusions & Relevance: Daily symptom screening and ACSs for sheltered homeless adults will substantially decrease COVID-19 cases and reduce costs compared to no intervention. In a surging epidemic, adding universal PCR testing every 2 weeks further decreases cases at modest incremental cost and should be considered. Keywords: Homelessness, COVID-19, cost-effectiveness analysis, simulation model

Clinical Impact, Costs, and Cost-Effectiveness of Expanded SARS-CoV-2 Testing in Massachusetts (preprint)

BackgroundWe projected the clinical and economic impact of alternative testing strategies on COVID-19 incidence and mortality in Massachusetts using a microsimulation model. MethodsWe compared five testing strategies: 1) PCR-severe-only: PCR testing only patients with severe/critical symptoms; 2) Self-screen: PCR-severe-only plus self-assessment of COVID-19-consistent symptoms with self-isolation if positive; 3) PCR-any-symptom: PCR for any COVID-19-consistent symptoms with self-isolation if positive; 4) PCR-all: PCR-any-symptom and one-time PCR for the entire population; and, 5) PCR-all-repeat: PCR-all with monthly re-testing. We examined effective reproduction numbers (Re, 0.9-2.0) at which policy conclusions would change. We used published data on disease progression and mortality, transmission, PCR sensitivity/specificity (70/100%) and costs. Model-projected outcomes included infections, deaths, tests performed, hospital-days, and costs over 180-days, as well as incremental cost-effectiveness ratios (ICERs, $/quality-adjusted life-year [QALY]). ResultsIn all scenarios, PCR-all-repeat would lead to the best clinical outcomes and PCR-severe-only would lead to the worst; at Re 0.9, PCR-all-repeat vs. PCR-severe-only resulted in a 63% reduction in infections and a 44% reduction in deaths, but required >65-fold more tests/day with 4-fold higher costs. PCR-all-repeat had an ICER

Cost-effectiveness of public health strategies for COVID-19 epidemic control in South Africa (preprint)

Background Healthcare resource constraints in low and middle-income countries necessitate selection of cost-effective public health interventions to address COVID-19. Methods We developed a dynamic COVID-19 microsimulation model to evaluate clinical and economic outcomes and cost-effectiveness of epidemic control strategies in KwaZulu-Natal, South Africa. Interventions assessed were Healthcare Testing (HT), where diagnostic testing is performed only for those presenting to healthcare centres; Contact Tracing (CT) in households of cases; Isolation Centres (IC), for cases not requiring hospitalisation; community health worker-led Mass Symptom Screening and diagnostic testing for symptomatic individuals (MS); and Quarantine Centres (QC), for contacts who test negative. Given uncertainties about epidemic dynamics in South Africa, we evaluated two main epidemic scenarios over 360 days, with effective reproduction numbers (Re) of 1.5 and 1.2. We compared HT, HT+CT, HT+CT+IC, HT+CT+IC+MS, HT+CT+IC+QC, and HT+CT+IC+MS+QC, considering strategies with incremental cost-effectiveness ratio (ICER)

AIDCOV: An Interpretable Artificial Intelligence Model for Detection of COVID-19 from Chest Radiography Images (preprint)

As the Coronavirus Disease 2019 (COVID-19) pandemic continues to grow globally, testing to detect COVID-19 and isolating individuals who test positive remains to be the primary strategy for preventing community spread of the disease. The current gold standard method of testing for COVID-19 is the reverse transcription polymerase chain reaction (RT-PCR) test. The RT-PCR test, however, has an imperfect sensitivity (around 70%), is time-consuming and labor-intensive, and is in short supply, particularly in resource-limited countries. Therefore, automatic and accurate detection of COVID-19 using medical imaging modalities such as chest X-ray and Computed Tomography, which are more widely available and accessible, can be beneficial as an alternative diagnostic tool. We develop a novel hierarchical attention neural network model to classify chest radiography images as belonging to a person with either COVID-19, other infections, or no pneumonia (i.e., normal). We refer to this model as Artificial Intelligence for Detection of COVID-19 (AIDCOV). The hierarchical structure in AIDCOV captures the dependency of features and improves model performance while the attention mechanism makes the model interpretable and transparent. Using a publicly available dataset of 5801 chest images, we demonstrate that our model achieves a mean cross-validation accuracy of 97.8%. AIDCOV has a sensitivity of 99.3%, a specificity of 99.98%, and a positive predictive value of 99.6% in detecting COVID-19 from chest radiography images. AIDCOV can be used in conjunction with or instead of RT-PCR testing (where RT-PCR testing is unavailable) to detect and isolate individuals with COVID-19 and prevent onward transmission to the general population and healthcare workers.