Challenges of COVID-19 Case Forecasting in the US, 2020-2021

During the COVID-19 pandemic, forecasting COVID-19 trends to support planning and response was a priority for scientists and decision makers alike. In the United States, COVID-19 forecasting was coordinated by a large group of universities, companies, and government entities led by the Centers for Disease Control and Prevention and the US COVID-19 Forecast Hub (https://covid19forecasthub.org). We evaluated approximately 9.7 million forecasts of weekly state-level COVID-19 cases for predictions 1-4 weeks into the future submitted by 24 teams from August 2020 to December 2021. We assessed coverage of central prediction intervals and weighted interval scores (WIS), adjusting for missing forecasts relative to a baseline forecast, and used a Gaussian generalized estimating equation (GEE) model to evaluate differences in skill across epidemic phases that were defined by the effective reproduction number. Overall, we found high variation in skill across individual models, with ensemble-based forecasts outperforming other approaches. Forecast skill relative to the baseline was generally higher for larger jurisdictions (e.g., states compared to counties). Over time, forecasts generally performed worst in periods of rapid changes in reported cases (either in increasing or decreasing epidemic phases) with 95% prediction interval coverage dropping below 50% during the growth phases of the winter 2020, Delta, and Omicron waves. Ideally, case forecasts could serve as a leading indicator of changes in transmission dynamics. However, while most COVID-19 case forecasts outperformed a naïve baseline model, even the most accurate case forecasts were unreliable in key phases. Further research could improve forecasts of leading indicators, like COVID-19 cases, by leveraging additional real-time data, addressing performance across phases, improving the characterization of forecast confidence, and ensuring that forecasts were coherent across spatial scales. In the meantime, it is critical for forecast users to appreciate current limitations and use a broad set of indicators to inform pandemic-related decision making.

Measuring effects of ivermectin-treated cattle on potential malaria vectors in Vietnam: A cluster-randomized trial

BACKGROUND

Malaria elimination using current tools has stalled in many areas. Ivermectin (IVM) is a broad-antiparasitic drug and mosquitocide and has been proposed as a tool for accelerating progress towards malaria elimination. Under laboratory conditions, IVM has been shown to reduce the survival of adult Anopheles populations that have fed on IVM-treated mammals. Treating cattle with IVM has been proposed as an important contribution to malaria vector management, however, the impacts of IVM in this One Health use case have been untested in field trials in Southeast Asia.

METHODS

Through a randomized village-based trial, this study quantified the effect of IVM-treated cattle on anopheline populations in treated vs. untreated villages in Central Vietnam. Local zebu cattle in six rural villages were included in this study. In three villages, cattle were treated with IVM at established veterinary dosages, and in three additional villages cattle were left as untreated controls. For the main study outcome, the mosquito populations in all villages were sampled using cattle-baited traps for six nights before, and six nights after a 2-day treatment IVM-administration (intervention) period. Anopheline species were characterized using taxonomic keys. The impact of the intervention was analyzed using a difference-in-differences (DID) approach with generalized estimating equations (with negative binomial distribution and robust errors). This intervention was powered to detect a 50% reduction in total nightly Anopheles spp. vector catches from cattle-baited traps. Given the unusual diversity in anopheline populations, exploratory analyses examined taxon-level differences in the ecological population diversity.

RESULTS

Across the treated villages, 1,112 of 1,527 censused cows (73% overall; range 67% to 83%) were treated with IVM. In both control and treated villages, there was a 30% to 40% decrease in total anophelines captured in the post-intervention period as compared to the pre-intervention period. In the control villages, there were 1,873 captured pre-intervention and 1,079 captured during the post-intervention period. In the treated villages, there were 1,594 captured pre-intervention, and 1,101 captured during the post-intervention period. The difference in differences model analysis comparing total captures between arms was not statistically significant (p = 0.61). Secondary outcomes of vector population diversity found that in three villages (one control and two treatment) Brillouin's index increased, and in three villages (two control and one treatment) Brillouin's index decreased. When examining biodiversity by trapping-night, there were no clear trends in treated or untreated vector populations. Additionally, there were no clear trends when examining the components of biodiversity: richness and evenness.

CONCLUSIONS

The ability of this study to quantify the impacts of IVM treatment was limited due to unexpectedly large spatiotemporal variability in trapping rates; an area-wide decrease in trapping counts across all six villages post-intervention; and potential spillover effects. However, this study provides important data to directly inform future studies in the GMS and beyond for IVM-based vector control.

Evaluation of FluSight influenza forecasting in the 2021-22 and 2022-23 seasons with a new target laboratory-confirmed influenza hospitalizations

Accurate forecasts can enable more effective public health responses during seasonal influenza epidemics. Forecasting teams were asked to provide national and jurisdiction-specific probabilistic predictions of weekly confirmed influenza hospital admissions for one through four weeks ahead for the 2021-22 and 2022-23 influenza seasons. Across both seasons, 26 teams submitted forecasts, with the submitting teams varying between seasons. Forecast skill was evaluated using the Weighted Interval Score (WIS), relative WIS, and coverage. Six out of 23 models outperformed the baseline model across forecast weeks and locations in 2021-22 and 12 out of 18 models in 2022-23. Averaging across all forecast targets, the FluSight ensemble was the 2nd most accurate model measured by WIS in 2021-22 and the 5th most accurate in the 2022-23 season. Forecast skill and 95% coverage for the FluSight ensemble and most component models degraded over longer forecast horizons and during periods of rapid change. Current influenza forecasting efforts help inform situational awareness, but research is needed to address limitations, including decreased performance during periods of changing epidemic dynamics.

Assessing the utility of COVID-19 case reports as a leading indicator for hospitalization forecasting in the United States

Identifying data streams that can consistently improve the accuracy of epidemiological forecasting models is challenging. Using models designed to predict daily state-level hospital admissions due to COVID-19 in California and Massachusetts, we investigated whether incorporating COVID-19 case data systematically improved forecast accuracy. Additionally, we considered whether using case data aggregated by date of test or by date of report from a surveillance system made a difference to the forecast accuracy. Evaluating forecast accuracy in a test period, after first having selected the best-performing methods in a validation period, we found that overall the difference in accuracy between approaches was small, especially at forecast horizons of less than two weeks. However, forecasts from models using cases aggregated by test date showed lower accuracy at longer horizons and at key moments in the pandemic, such as the peak of the Omicron wave in January 2022. Overall, these results highlight the challenge of finding a modeling approach that can generate accurate forecasts of outbreak trends both during periods of relative stability and during periods that show rapid growth or decay of transmission rates. While COVID-19 case counts seem to be a natural choice to help predict COVID-19 hospitalizations, in practice any benefits we observed were small and inconsistent.

Pharmacogenomic variation in the Malagasy population: implications for the antimalarial drug primaquine metabolism

Aim: Antimalarial primaquine (PQ) eliminates liver hypnozoites of Plasmodium vivax. CYP2D6 gene variation contributes to PQ therapeutic failure. Additional gene variation may contribute to PQ efficacy. Information on pharmacogenomic variation in Madagascar, with vivax malaria and a unique population admixture, is scanty. Methods: The authors performed genome-wide genotyping of 55 Malagasy samples and analyzed data with a focus on a set of 28 pharmacogenes most relevant to PQ. Results: Mainly, the study identified 110 coding or splicing variants, including those that, based on previous studies in other populations, may be implicated in PQ response and copy number variation, specifically in chromosomal regions that contain pharmacogenes. Conclusion: With this pilot information, larger genome-wide association analyses with PQ metabolism and response are substantially more feasible.

Assessing the utility of COVID-19 case reports as a leading indicator for hospitalization forecasting in the United States

Identifying data streams that can consistently improve the accuracy of epidemiological forecasting models is challenging. Using models designed to predict daily state-level hospital admissions due to COVID-19 in California and Massachusetts, we investigated whether incorporating COVID-19 case data systematically improved forecast accuracy. Additionally, we considered whether using case data aggregated by date of test or by date of report from a surveillance system made a difference to the forecast accuracy. Evaluating forecast accuracy in a test period, after first having selected the best-performing methods in a validation period, we found that overall the difference in accuracy between approaches was small, especially at forecast horizons of less than two weeks. However, forecasts from models using cases aggregated by test date showed lower accuracy at longer horizons and at key moments in the pandemic, such as the peak of the Omicron wave in January 2022. Overall, these results highlight the challenge of finding a modeling approach that can generate accurate forecasts of outbreak trends both during periods of relative stability and during periods that show rapid growth or decay of transmission rates. While COVID-19 case counts seem to be a natural choice to help predict COVID-19 hospitalizations, in practice any benefits we observed were small and inconsistent.

Optimizing the implementation of a participant-collected, mail-based SARS-CoV-2 serological survey in university-affiliated populations: lessons learned and practical guidance

The rapid spread of SARS-CoV-2 is largely driven by pre-symptomatic or mildly symptomatic individuals transmitting the virus. Serological tests to identify antibodies against SARS-CoV-2 are important tools to characterize subclinical infection exposure.

During the summer of 2020, a mail-based serological survey with self-collected dried blood spot (DBS) samples was implemented among university affiliates and their household members in Massachusetts, USA. Described are challenges faced and novel procedures used during the implementation of this study to assess the prevalence of SARS-CoV-2 antibodies amid the pandemic.

Important challenges included user-friendly remote and contact-minimized participant recruitment, limited availability of some commodities and laboratory capacity, a potentially biased sample population, and policy changes impacting the distribution of clinical results to study participants. Methods and lessons learned to surmount these challenges are presented to inform design and implementation of similar sero-studies.

This study design highlights the feasibility and acceptability of self-collected bio-samples and has broad applicability for other serological surveys for a range of pathogens. Key lessons relate to DBS sampling, supply requirements, the logistics of packing and shipping packages, data linkages to enrolled household members, and the utility of having an on-call nurse available for participant concerns during sample collection. Future research might consider additional recruitment techniques such as conducting studies during academic semesters when recruiting in a university setting, partnerships with supply and shipping specialists, and using a stratified sampling approach to minimize potential biases in recruitment.

The United States COVID-19 Forecast Hub dataset

Academic researchers, government agencies, industry groups, and individuals have produced forecasts at an unprecedented scale during the COVID-19 pandemic. To leverage these forecasts, the United States Centers for Disease Control and Prevention (CDC) partnered with an academic research lab at the University of Massachusetts Amherst to create the US COVID-19 Forecast Hub. Launched in April 2020, the Forecast Hub is a dataset with point and probabilistic forecasts of incident cases, incident hospitalizations, incident deaths, and cumulative deaths due to COVID-19 at county, state, and national, levels in the United States. Included forecasts represent a variety of modeling approaches, data sources, and assumptions regarding the spread of COVID-19. The goal of this dataset is to establish a standardized and comparable set of short-term forecasts from modeling teams. These data can be used to develop ensemble models, communicate forecasts to the public, create visualizations, compare models, and inform policies regarding COVID-19 mitigation. These open-source data are available via download from GitHub, through an online API, and through R packages.

The United States COVID-19 Forecast Hub dataset

Academic researchers, government agencies, industry groups, and individuals have produced forecasts at an unprecedented scale during the COVID-19 pandemic. To leverage these forecasts, the United States Centers for Disease Control and Prevention (CDC) partnered with an academic research lab at the University of Massachusetts Amherst to create the US COVID-19 Forecast Hub. Launched in April 2020, the Forecast Hub is a dataset with point and probabilistic forecasts of incident cases, incident hospitalizations, incident deaths, and cumulative deaths due to COVID-19 at county, state, and national, levels in the United States. Included forecasts represent a variety of modeling approaches, data sources, and assumptions regarding the spread of COVID-19. The goal of this dataset is to establish a standardized and comparable set of short-term forecasts from modeling teams. These data can be used to develop ensemble models, communicate forecasts to the public, create visualizations, compare models, and inform policies regarding COVID-19 mitigation. These open-source data are available via download from GitHub, through an online API, and through R packages.

Comparing trained and untrained probabilistic ensemble forecasts of COVID-19 cases and deaths in the United States

The U.S. COVID-19 Forecast Hub aggregates forecasts of the short-term burden of COVID-19 in the United States from many contributing teams. We study methods for building an ensemble that combines forecasts from these teams. These experiments have informed the ensemble methods used by the Hub. To be most useful to policy makers, ensemble forecasts must have stable performance in the presence of two key characteristics of the component forecasts: (1) occasional misalignment with the reported data, and (2) instability in the relative performance of component forecasters over time. Our results indicate that in the presence of these challenges, an untrained and robust approach to ensembling using an equally weighted median of all component forecasts is a good choice to support public health decision makers. In settings where some contributing forecasters have a stable record of good performance, trained ensembles that give those forecasters higher weight can also be helpful.

Evaluation of individual and ensemble probabilistic forecasts of COVID-19 mortality in the United States

Short-term probabilistic forecasts of the trajectory of the COVID-19 pandemic in the United States have served as a visible and important communication channel between the scientific modeling community and both the general public and decision-makers. Forecasting models provide specific, quantitative, and evaluable predictions that inform short-term decisions such as healthcare staffing needs, school closures, and allocation of medical supplies. Starting in April 2020, the US COVID-19 Forecast Hub (https://covid19forecasthub.org/) collected, disseminated, and synthesized tens of millions of specific predictions from more than 90 different academic, industry, and independent research groups. A multimodel ensemble forecast that combined predictions from dozens of groups every week provided the most consistently accurate probabilistic forecasts of incident deaths due to COVID-19 at the state and national level from April 2020 through October 2021. The performance of 27 individual models that submitted complete forecasts of COVID-19 deaths consistently throughout this year showed high variability in forecast skill across time, geospatial units, and forecast horizons. Two-thirds of the models evaluated showed better accuracy than a naïve baseline model. Forecast accuracy degraded as models made predictions further into the future, with probabilistic error at a 20-wk horizon three to five times larger than when predicting at a 1-wk horizon. This project underscores the role that collaboration and active coordination between governmental public-health agencies, academic modeling teams, and industry partners can play in developing modern modeling capabilities to support local, state, and federal response to outbreaks.

Serological surveys to estimate cumulative incidence of SARS-CoV-2 infection in adults (Sero-MAss study), Massachusetts, July-August 2020: a mail-based cross-sectional study

OBJECTIVES

To estimate the seroprevalence of anti-SARS-CoV-2 IgG and IgM among Massachusetts residents and to better understand asymptomatic SARS-CoV-2 transmission during the summer of 2020.

DESIGN

Mail-based cross-sectional survey.

SETTING

Massachusetts, USA.

PARTICIPANTS

Primary sampling group: sample of undergraduate students at the University of Massachusetts, Amherst (n=548) and a member of their household (n=231).Secondary sampling group: sample of graduate students, faculty, librarians and staff (n=214) and one member of their household (n=78). All participants were residents of Massachusetts without prior COVID-19 diagnosis.

PRIMARY AND SECONDARY OUTCOME MEASURES

Prevalence of SARS-CoV-2 seropositivity. Association of seroprevalence with variables including age, gender, race, geographic region, occupation and symptoms.

RESULTS

Approximately 27 000 persons were invited via email to assess eligibility. 1001 households were mailed dried blood spot sample kits, 762 returned blood samples for analysis. In the primary sample group, 36 individuals (4.6%) had IgG antibodies detected for an estimated weighted prevalence in this population of 5.3% (95% CI: 3.5 to 8.0). In the secondary sampling group, 10 participants (3.4%) had IgG antibodies detected for an estimated adjusted prevalence of 4.0% (95% CI: 2.2 to 7.4). No samples were IgM positive. No association was found in either group between seropositivity and self-reported work duties or customer-facing hours. In the primary sampling group, self-reported febrile illness since February 2020, male sex and minority race (Black or American Indian/Alaskan Native) were associated with seropositivity. No factors except geographic regions within the state were associated with evidence of prior SARS-CoV-2 infection in the secondary sampling group.

CONCLUSIONS

This study fills a critical gap in estimating the levels of subclinical and asymptomatic infection. Estimates can be used to calibrate models estimating levels of population immunity over time, and these data are critical for informing public health interventions and policy.

Serological surveys to estimate cumulative incidence of SARS-CoV-2 infection in adults (Sero-MAss study), Massachusetts, July-August 2020: a mail-based cross-sectional study

BACKGROUND

The SARS-CoV-2 pandemic is an unprecedented global health crisis. The state of Massachusetts was especially impacted during the initial stages; however, the extent of asymptomatic transmission remains poorly understood due to limited asymptomatic testing in the "first wave." To address this gap, a geographically representative and contact-free seroprevalence survey was conducted in July-August 2020, to estimate prior undetected SARS-CoV-2 infections.

METHODS

Students, faculty, librarians and staff members at the University of Massachusetts, Amherst without a previous COVID-19 diagnosis were invited to participate in this study along with one member of their household in June 2020. Two separate sampling frames were generated from administrative lists: all undergraduates and their household members (primary sampling group) were randomly selected with probability proportional to population size. All staff, faculty, graduate students and librarians (secondary sampling group) were selected as a simple random sample. After informed consent and a socio-behavioral survey, participants were mailed test kits and asked to return self-collected dried blood spot (DBS) samples. Samples were analyzed via ELISA for anti-SARS-CoV-2 IgG antibodies, and then IgM antibodies if IgG-positive. Seroprevalence estimates were adjusted for survey non-response. Binomial models were used to assess factors associated with seropositivity in both sample groups separately.

RESULTS

Approximately 27,000 persons were invited via email to assess eligibility. Of the 1,001 individuals invited to participate in the study, 762 (76%) returned blood samples for analysis. In the primary sampling group 548 returned samples, of which 230 enrolled a household member. Within the secondary sampling group of 214 individuals, 79 enrolled a household member. In the primary sample group, 36 (4.6%) had IgG antibodies detected for an estimated weighed prevalence for this population of 5.3% (95% CI: 3.5 to 8.0). In the secondary sampling group, 10 (3.4%) of 292 individuals had IgG antibodies detected for an estimated adjusted prevalence of 4.0% (95% CI: 2.2 to 7.4). No samples were IgM positive. No association was found in either sample group between seropositivity and self-reported work duties or customer-facing hours. In the primary sampling group, self-reported febrile illness since Feb 2020, male sex, and minority race (Black or American Indian/Alaskan Native) were associated with seropositivity. No factors except geographic regions within the state were associated with evidence of prior SARS-CoV-2 infection in the secondary sampling group.

INTERPRETATION

This study provides insight into the seroprevalence of university-related populations and their household members across the state of Massachusetts during the summer of 2020 of the pandemic and helps to fill a critical gap in estimating the levels of sub-clinical and asymptomatic infection. Estimates like these can be used to calibrate models that estimate levels of population immunity over time to inform public health interventions and policy.

Ivermectin Treatment for Cattle Reduced the Survival of Two Malaria Vectors, Anopheles dirus and Anopheles epiroticus, Under Laboratory Conditions in Central Vietnam

Ivermectin is a low-cost and nontoxic mosquitocide that may have a role in malaria elimination. However, the extent to which this drug impacts the mortality of Anopheles dirus and Anopheles epiroticus, two important malaria vectors in Southeast Asia, is unknown. This study compared and quantified anopheline mortality after feeding on ivermectin-treated cattle and control cattle in Vietnam. Local anopheline colonies fed on cattle 1 to 3, 6 to 8, 13 to 15, 20 to 22, and 28 to 30 days after injection (DAI) with ivermectin (intervention) or saline (control). An. dirus that fed on ivermectin-treated cattle had higher mortality rates than controls for up to 20 DAI (P < 0.05); An. epiroticus that fed on ivermectin-treated cattle had consistently higher mortality rates than controls for up to 8 DAI (P < 0.05). Feeding on ivermectin-treated cattle increased the mortality rate of these vector species for biologically relevant time periods. Therefore, ivermectin has the potential to become an important tool for integrated vector management.

Plasmodium falciparum Parasitemia and Band Sensitivity of the SD Bioline Malaria Ag P.f/Pan Rapid Diagnostic Test in Madagascar

Current malaria rapid diagnostic tests (RDTs) contain antibodies against Plasmodium falciparum-specific histidine-rich protein 2 (PfHRP2), Plasmodium lactate dehydrogenase (pLDH), and aldolase in various combinations. Low or high parasite densities/target antigen concentrations may influence the accuracy and sensitivity of PfHRP2-detecting RDTs. We analyzed the SD Bioline Malaria Ag P.f/Pan RDT performance in relation to P. falciparum parasitemia in Madagascar, where clinical Plasmodium vivax malaria exists alongside P. falciparum. Nine hundred sixty-three samples from patients seeking care for suspected malaria infection were analyzed by RDT, microscopy, and Plasmodium species-specific, ligase detection reaction-fluorescent microsphere assay (LDR-FMA). Plasmodium infection positivity by these diagnostics was 47.9%, 46.9%, and 58%, respectively. Plasmodium falciparum-only infections were predominant (microscopy, 45.7%; LDR-FMA, 52.3%). In all, 16.3% of P. falciparum, 70% of P. vivax, and all of Plasmodium malariae, Plasmodium ovale, and mixed-species infections were submicroscopic. In 423 P. falciparum mono-infections, confirmed by microscopy and LDR-FMA, the parasitemia in those who were positive for both the PfHRP2 and pan-pLDH test bands was significantly higher than that in those who were positive only for the PfHRP2 band (P < 0.0001). Plasmodium falciparum parasitemia in those that were detected as P. falciparum-only infections by microscopy but P. falciparum mixed infections by LDR-FMA also showed similar outcome by the RDT band positivity. In addition, we used varying parasitemia (3-0.0001%) of the laboratory-maintained 3D7 strain to validate this observation. A positive pLDH band in high P. falciparum-parasitemic individuals may complicate diagnosis and treatment, particularly when the microscopy is inconclusive for P. vivax, and the two infections require different treatments.

Risk Factors for Malaria Infection in Central Madagascar: Insights from a Cross-Sectional Population Survey

Community prevalence of infection is a widely used, standardized metric for evaluating malaria endemicity. Conventional methods for measuring prevalence include light microscopy and rapid diagnostic tests (RDTs), but their detection thresholds are inadequate for diagnosing low-density infections. The significance of submicroscopic malaria infections is poorly understood in Madagascar, a country of heterogeneous malaria epidemiology. A cross-sectional community survey in the western foothills of Madagascar during the March 2014 transmission season found malaria infection to be predominantly submicroscopic and asymptomatic. Prevalence of Plasmodium infection diagnosed by microscopy, RDT, and molecular diagnosis was 2.4%, 4.1%, and 13.8%, respectively. This diagnostic discordance was greatest for Plasmodium vivax infection, which was 98.5% submicroscopic. Village location, insecticide-treated bednet ownership, and fever were significantly associated with infection outcomes, as was presence of another infected individual in the household. Duffy-negative individuals were diagnosed with P. vivax, but with reduced odds relative to Duffy-positive hosts. The observation of high proportions of submicroscopic infections calls for a wider assessment of the parasite reservoir in other regions of the island, particularly given the country's current focus on malaria elimination and the poorly documented distribution of the non-P. falciparum parasite species.