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.

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.

Can severity of a humanitarian crisis be quantified? Assessment of the INFORM severity index.

BACKGROUND: Those responding to humanitarian crises have an ethical imperative to respond most where the need is greatest. Metrics are used to estimate the severity of a given crisis. The INFORM Severity Index, one such metric, has become widely used to guide policy makers in humanitarian response decision making. The index, however, has not undergone critical statistical review. If imprecise or incorrect, the quality of decision making for humanitarian response will be affected. This analysis asks, how precise and how well does this index reflect the severity of conditions for people affected by disaster or war? RESULTS: The INFORM Severity Index is calculated from 35 publicly available indicators, which conceptually reflect the severity of each crisis. We used 172 unique global crises from the INFORM Severity Index database that occurred January 1 to November 30, 2019 or were ongoing by this date. We applied exploratory factor analysis (EFA) to determine common factors within the dataset. We then applied a second-order confirmatory factor analysis (CFA) to predict crisis severity as a latent construct. Model fit was assessed via chi-square goodness-of-fit statistic, Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), and Root Mean Square Error of Approximation (RMSEA). The EFA models suggested a 3- or 4- factor solution, with 46 and 53% variance explained in each model, respectively. The final CFA was parsimonious, containing three factors comprised of 11 indicators, with reasonable model fit (Chi-squared = 107, with 40 degrees of freedom, CFI = 0.94, TLI = 0.92, RMSEA = 0.10). In the second-order CFA, the magnitude of standardized factor-loading on the 'societal governance' latent construct had the strongest association with the latent construct of 'crisis severity' (0.73), followed by the 'humanitarian access/safety' construct (0.56). CONCLUSIONS: A metric of crisis-severity is a critical step towards improving humanitarian response, but only when it reflects real life conditions. Our work is a first step in refining an existing framework to better quantify crisis severity.

Trends of child undernutrition in rural Ecuadorian communities with differential access to roads, 2004-2013.

Road access can influence protective and risk factors associated with nutrition by affecting various social and biological processes. In northern coastal Ecuador, the construction of new roads created a remoteness gradient among villages, providing a unique opportunity to examine the impact of roads on child nutritional outcomes 10 years after the road was built. Anthropometric and haemoglobin measurements were collected from 2,350 children <5 years in Esmeraldas, Ecuador, from 2004 to 2013 across 28 villages with differing road access. Logistic generalized estimating equation models assessed the longitudinal association between village remoteness and prevalence of stunting, wasting, underweight, overweight, obesity, and anaemia. We examined the influence of socio-economic characteristics on the pathway between remoteness and nutrition by comparing model results with and without household-level socio-economic covariates. Remoteness was associated with stunting (OR = 0.43, 95% CI [0.30, 0.63]) and anaemia (OR = 0.56, 95% CI [0.44, 0.70]). Over time, the prevalence of stunting was generally decreasing but remained higher in villages closer to the road compared to those farther away. Obesity increased (0.5% to 3%) over time; wasting was high (6%) but stable during the study period. Wealth and education partially explained the better nutritional outcomes in remote vs. road villages more than a decade after some communities gained road access. Establishing the extent to which these patterns persist requires additional years of observation.

Lessons learned from implementation of a national hotline for Ebola virus disease emergency preparedness in South Sudan.

BACKGROUND: The world's second largest Ebola outbreak occurred in the Democratic Republic of Congo from 2018 to 2020. At the time, risk of cross-border spread into South Sudan was very high. Thus, the South Sudan Ministry of Health scaled up Ebola preparedness activities in August 2018, including implementation of a 24-h, toll-free Ebola virus disease (EVD) hotline. The primary purpose was the hotline was to receive EVD alerts and the secondary goal was to provide evidence-based EVD messages to the public. METHODS: To assess whether the hotline augmented Ebola preparedness activities in a protracted humanitarian emergency context, we reviewed 22 weeks of call logs from January to June 2019. Counts and percentages were calculated for all available data. RESULTS: The hotline received 2114 calls during the analysis period, and an additional 1835 missed calls were documented. Callers used the hotline throughout 24-h of the day and were most often men and individuals living in Jubek state, where the national capital is located. The leading reasons for calling were to learn more about EVD (68%) or to report clinical signs or symptoms (16%). Common EVD-related questions included EVD signs and symptoms, transmission, and prevention. Only one call was documented as an EVD alert, and there was no documentation of reported symptoms or whether the person met the EVD case definition. CONCLUSIONS: Basic surveillance information was not collected from callers. To trigger effective outbreak investigation from hotline calls, the hotline should capture who is reporting and from where, symptoms and travel history, and whether this information should be further investigated. Electronic data capture will enhance data quality and availability of information for review. Additionally, the magnitude of missed calls presents a major challenge. When calls are answered, there is potential to provide health communication, so risk communication needs should be considered. However, prior to hotline implementation, governments should critically assess whether their hotline would yield actionable data and if other data sources for surveillance or community concerns are available.

Breastfeeding support in low and middle-income countries: Secondary analysis of national survey data.

BACKGROUND: Early initiation of breastfeeding and exclusive breastfeeding can reduce infant mortality. Breastfeeding support interventions such as counselling may improve adherence to recommended practices. However, it is not known if these interventions work at the population level. OBJECTIVE: The aim of this study was to assess the relationship between early postnatal breastfeeding support and recommended breastfeeding practices. DESIGN/SETTING: We pooled data from 11 Demographic and Health Surveys in Africa (n = 7), South East Asia (n = 2), the Americas (n = 1), and Europe (n = 1) to analyse these associations at the population level. PARTICIPANTS: We limited the data to the most recent live births in the two years before the survey, including 41,431 births. ANALYSIS: We fitted three multivariable logistic regression models to estimate the relationship between early postnatal breastfeeding support (a newborn postnatal check within an hour of birth plus counselling and observation of breastfeeding within two days) and three breastfeeding outcomes (early initiation of breastfeeding, absence of prelacteal feeding, and exclusive breastfeeding), adjusting for sociodemographic characteristics and birth-related factors. FINDINGS: Early breastfeeding support was associated with a 24% increase (OR=1.24 95%CI=1.11,1.39) in the odds of initiating breastfeeding within one hour of birth. No relationships were found between breastfeeding support and prelacteal feeding in the first three days or exclusive breastfeeding at six months. KEY CONCLUSION: While postnatal breastfeeding counselling and observation may improve early initiation of breastfeeding, impact is not persistent for longer term breastfeeding outcomes. IMPLICATION FOR PRACTICE: Improved training for breastfeeding support and an enabling policy environment are required to improve breastfeeding practices for women and newborns.

Determinants of Latrine Use Behavior: The Psychosocial Proxies of Individual-Level Defecation Practices in Rural Coastal Ecuador.

There is increasing appreciation that latrine access does not imply use-many individuals who own latrines do not consistently use them. Little is known, however, about the determinants of latrine use, particularly among those with variable defecation behaviors. Using the integrated behavior model of water, sanitation, and hygiene framework, we sought to characterize determinants of latrine use in rural Ecuador. We interviewed 197 adults living in three communities with a survey consisting of 70 psychosocial defecation-related questions. Questions were excluded from analysis if responses lacked variability or at least 10% of respondents did not provide a definitive answer. All interviewed individuals had access to a privately owned or shared latrine. We then applied adaptive elastic nets (ENET) and supervised principal component analysis (SPCA) to a reduced dataset of 45 questions among 154 individuals with complete data to select determinants that predict self-reported latrine use. Latrine use was common, but not universal, in the sample (76%). The SPCA model identified six determinants and adaptive ENET selected five determinants. Three indicators were represented in both models-latrine users were more likely to report that their latrine is clean enough to use and also more likely to report daily latrine use; while those reporting that elderly men were not latrine users were less likely to use latrines themselves. Our findings suggest that social norms are important predictors of latrine use, whereas knowledge of the health benefits of sanitation may not be as important. These determinants are informative for promotion of latrine adoption.

Distribution of Enteroinvasive and Enterotoxigenic Escherichia coli Across Space and Time in Northwestern Ecuador.

Although Escherichia coli infections are common throughout the developing world, their prevalence patterns in space and over time are not well characterized. We used serial case control data collected from 16 communities in northwestern Ecuador between 2004 and 2010, to examine the prevalence of enteroinvasive E. coli (EIEC) and enterotoxigenic E. coli (ETEC). At its peak, the regional prevalence of EIEC was 8.3 infections/100 persons but this decreased to 1 infection/1,000 persons. The regional prevalence of ETEC ranged from 8 infections/1,000 persons to 3.7 infections/100 persons. The prevalence pattern of EIEC resembled that of a large epidemic whereas the prevalence of ETEC was more stable over time. Here, we provide community-based evidence for temporal shifts in the dominant E. coli pathotype from EIEC to ETEC over a multi-year time period. Furthermore, genotype analysis suggests that a given strain of EIEC and ETEC can persist in this region for long periods, up to 24 and 55 months, respectively.