Elaine O. Nsoesie

Combining Search, Social Media, and Traditional Data Sources to Improve Influenza Surveillance

We present a machine learning-based methodology capable of providing real-time (“nowcast”) and forecast estimates of influenza activity in the US by leveraging data from multiple data sources including: Google searches, Twitter microblogs, nearly real-time hospital visit records, and data from a participatory surveillance system. Our main contribution consists of combining multiple influenza-like illnesses (ILI) activity estimates, generated independently with each data source, into a single prediction of ILI utilizing machine learning ensemble approaches. Our methodology exploits the information in each data source and produces accurate weekly ILI predictions for up to four weeks ahead of the release of CDC’s ILI reports. We evaluate the predictive ability of our ensemble approach during the 2013–2014 (retrospective) and 2014–2015 (live) flu seasons for each of the four weekly time horizons. Our ensemble approach demonstrates several advantages: (1) our ensemble method’s predictions outperform every prediction using each data source independently, (2) our methodology can produce predictions one week ahead of GFT’s real-time estimates with comparable accuracy, and (3) our two and three week forecast estimates have comparable accuracy to real-time predictions using an autoregressive model. Moreover, our results show that considerable insight is gained from incorporating disparate data streams, in the form of social media and crowd sourced data, into influenza predictions in all time horizons.

Monitoring Influenza Epidemics in China with Search Query from Baidu

Several approaches have been proposed for near real-time detection and prediction of the spread of influenza. These include search query data for influenza-related terms, which has been explored as a tool for augmenting traditional surveillance methods. In this paper, we present a method that uses Internet search query data from Baidu to model and monitor influenza activity in China. The objectives of the study are to present a comprehensive technique for: (i) keyword selection, (ii) keyword filtering, (iii) index composition and (iv) modeling and detection of influenza activity in China. Sequential time-series for the selected composite keyword index is significantly correlated with Chinese influenza case data. In addition, one-month ahead prediction of influenza cases for the first eight months of 2012 has a mean absolute percent error less than 11%. To our knowledge, this is the first study on the use of search query data from Baidu in conjunction with this approach for estimation of influenza activity in China.

A systematic review of studies on forecasting the dynamics of influenza outbreaks

Forecasting the dynamics of influenza outbreaks could be useful for decision‐making regarding the allocation of public health resources. Reliable forecasts could also aid in the selection and implementation of interventions to reduce morbidity and mortality due to influenza illness. This paper reviews methods for influenza forecasting proposed during previous influenza outbreaks and those evaluated in hindsight. We discuss the various approaches, in addition to the variability in measures of accuracy and precision of predicted measures. PubMed and Google Scholar searches for articles on influenza forecasting retrieved sixteen studies that matched the study criteria. We focused on studies that aimed at forecasting influenza outbreaks at the local, regional, national, or global level. The selected studies spanned a wide range of regions including USA, Sweden, Hong Kong, Japan, Singapore, United Kingdom, Canada, France, and Cuba. The methods were also applied to forecast a single measure or multiple measures. Typical measures predicted included peak timing, peak height, daily/weekly case counts, and outbreak magnitude. Due to differences in measures used to assess accuracy, a single estimate of predictive error for each of the measures was difficult to obtain. However, collectively, the results suggest that these diverse approaches to influenza forecasting are capable of capturing specific outbreak measures with some degree of accuracy given reliable data and correct disease assumptions. Nonetheless, several of these approaches need to be evaluated and their performance quantified in real‐time predictions.

Enhancing disease surveillance with novel data streams: challenges and opportunities

Novel data streams (NDS), such as web search data or social media updates, hold promise for enhancing the capabilities of public health surveillance. In this paper, we outline a conceptual framework for integrating NDS into current public health surveillance. Our approach focuses on two key questions: What are the opportunities for using NDS and what are the minimal tests of validity and utility that must be applied when using NDS? Identifying these opportunities will necessitate the involvement of public health authorities and an appreciation of the diversity of objectives and scales across agencies at different levels (local, state, national, international). We present the case that clearly articulating surveillance objectives and systematically evaluating NDS and comparing the performance of NDS to existing surveillance data and alternative NDS data is critical and has not sufficiently been addressed in many applications of NDS currently in the literature.

Assessing the origin of and potential for international spread of Chikungunya virus from the Caribbean

Background: For the first time, an outbreak of chikungunya has been reported in the Americas. Locally acquired infections have been confirmed in fourteen Caribbean countries and dependent territories, Guyana and French Guiana, in which a large number of North American travelers vacation. Should some travelers become infected with chikungunya virus, they could potentially introduce it into the United States, where there are competent Aedes mosquito vectors, with the possibility of local transmission. Methods: We analyzed historical data on airline travelers departing areas of the Caribbean and South America, where locally acquired cases of chikungunya have been confirmed as of May 12th, 2014. The final destinations of travelers departing these areas between May and July 2012 were determined and overlaid on maps of the reported distribution of Aedes aeygpti and albopictus mosquitoes in the United States, to identify potential areas at risk of autochthonous transmission. Results: The United States alone accounted for 52.1% of the final destinations of all international travelers departing chikungunya indigenous areas of the Caribbean between May and July 2012. Cities in the United States with the highest volume of air travelers were New York City, Miami and San Juan (Puerto Rico). Miami and San Juan were high travel-volume cities where Aedes aeygpti or albopictus are reported and where climatic conditions could be suitable for autochthonous transmission. Conclusion: The rapidly evolving outbreak of chikungunya in the Caribbean poses a growing risk to countries and areas linked by air travel, including the United States where competent Aedes mosquitoes exist. The risk of chikungunya importation into the United States may be elevated following key travel periods in the spring, when large numbers of North American travelers typically vacation in the Caribbean.

Spread of yellow fever virus outbreak in Angola and the Democratic Republic of the Congo 2015–16: a modelling study

Summary Background Since late 2015, an epidemic of yellow fever has caused more than 7334 suspected cases in Angola and the Democratic Republic of the Congo, including 393 deaths. We sought to understand the spatial spread of this outbreak to optimise the use of the limited available vaccine stock. Methods We jointly analysed datasets describing the epidemic of yellow fever, vector suitability, human demography, and mobility in central Africa to understand and predict the spread of yellow fever virus. We used a standard logistic model to infer the district-specific yellow fever virus infection risk during the course of the epidemic in the region. Findings The early spread of yellow fever virus was characterised by fast exponential growth (doubling time of 5–7 days) and fast spatial expansion (49 districts reported cases after only 3 months) from Luanda, the capital of Angola. Early invasion was positively correlated with high population density (Pearsons r 0·52, 95% CI 0·34–0·66). The further away locations were from Luanda, the later the date of invasion (Pearsons r 0·60, 95% CI 0·52–0·66). In a Cox model, we noted that districts with higher population densities also had higher risks of sustained transmission (the hazard ratio for cases ceasing was 0·74, 95% CI 0·13–0·92 per log-unit increase in the population size of a district). A model that captured human mobility and vector suitability successfully discriminated districts with high risk of invasion from others with a lower risk (area under the curve 0·94, 95% CI 0·92–0·97). If at the start of the epidemic, sufficient vaccines had been available to target 50 out of 313 districts in the area, our model would have correctly identified 27 (84%) of the 32 districts that were eventually affected. Interpretation Our findings show the contributions of ecological and demographic factors to the ongoing spread of the yellow fever outbreak and provide estimates of the areas that could be prioritised for vaccination, although other constraints such as vaccine supply and delivery need to be accounted for before such insights can be translated into policy. Funding Wellcome Trust.

Measuring patient-perceived quality of care in US hospitals using Twitter

Background Patients routinely use Twitter to share feedback about their experience receiving healthcare. Identifying and analysing the content of posts sent to hospitals may provide a novel real-time measure of quality, supplementing traditional, survey-based approaches. Objective To assess the use of Twitter as a supplemental data stream for measuring patient-perceived quality of care in US hospitals and compare patient sentiments about hospitals with established quality measures. Design 404 065 tweets directed to 2349 US hospitals over a 1-year period were classified as having to do with patient experience using a machine learning approach. Sentiment was calculated for these tweets using natural language processing. 11 602 tweets were manually categorised into patient experience topics. Finally, hospitals with ≥50 patient experience tweets were surveyed to understand how they use Twitter to interact with patients. Key results Roughly half of the hospitals in the US have a presence on Twitter. Of the tweets directed toward these hospitals, 34 725 (9.4%) were related to patient experience and covered diverse topics. Analyses limited to hospitals with ≥50 patient experience tweets revealed that they were more active on Twitter, more likely to be below the national median of Medicare patients (p<0.001) and above the national median for nurse/patient ratio (p=0.006), and to be a non-profit hospital (p<0.001). After adjusting for hospital characteristics, we found that Twitter sentiment was not associated with Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) ratings (but having a Twitter account was), although there was a weak association with 30-day hospital readmission rates (p=0.003). Conclusions Tweets describing patient experiences in hospitals cover a wide range of patient care aspects and can be identified using automated approaches. These tweets represent a potentially untapped indicator of quality and may be valuable to patients, researchers, policy makers and hospital administrators.

Online reports of foodborne illness capture foods implicated in official foodborne outbreak reports

OBJECTIVE Traditional surveillance systems capture only a fraction of the estimated 48 million yearly cases of foodborne illness in the United States. We assessed whether foodservice reviews on Yelp.com (a business review site) can be used to support foodborne illness surveillance efforts. METHODS We obtained reviews from 2005 to 2012 of 5824 foodservice businesses closest to 29 colleges. After extracting recent reviews describing episodes of foodborne illness, we compared implicated foods to foods in outbreak reports from the U.S. Centers for Disease Control and Prevention (CDC). RESULTS Broadly, the distribution of implicated foods across five categories was as follows: aquatic (16% Yelp, 12% CDC), dairy-eggs (23% Yelp, 23% CDC), fruits-nuts (7% Yelp, 7% CDC), meat-poultry (32% Yelp, 33% CDC), and vegetables (22% Yelp, 25% CDC). The distribution of foods across 19 more specific food categories was also similar, with Spearman correlations ranging from 0.60 to 0.85 for 2006-2011. The most implicated food categories in both Yelp and CDC were beef, dairy, grains-beans, poultry and vine-stalk. CONCLUSIONS Based on observations in this study and the increased usage of social media, we posit that online illness reports could complement traditional surveillance systems by providing near real-time information on foodborne illnesses, implicated foods and locations.

Global skin disease morbidity and mortality an update from the global burden of disease study 2013

Importance Disability secondary to skin conditions is substantial worldwide. The Global Burden of Disease Study 2013 includes estimates of global morbidity and mortality due to skin diseases. Objective To measure the burden of skin diseases worldwide. Data Sources For nonfatal estimates, data were found by literature search using PubMed and Google Scholar in English and Spanish for years 1980 through 2013 and by accessing administrative data on hospital inpatient and outpatient episodes. Data for fatal estimates were based on vital registration and verbal autopsy data. Study Selection Skin disease data were extracted from more than 4000 sources including systematic reviews, surveys, population-based disease registries, hospital inpatient data, outpatient data, cohort studies, and autopsy data. Data metrics included incidence, prevalence, remission, duration, severity, deaths, and mortality risk. Data Extraction and Synthesis Data were extracted by age, time period, case definitions, and other study characteristics. Data points were modeled with Bayesian meta-regression to generate estimates of morbidity and mortality metrics for skin diseases. All estimates were made with 95% uncertainty intervals. Main Outcomes and Measures Disability-adjusted life years (DALYs), years lived with disability, and years of life lost from 15 skin conditions in 188 countries. Results Skin conditions contributed 1.79% to the global burden of disease measured in DALYs from 306 diseases and injuries in 2013. Individual skin diseases varied in size from 0.38% of total burden for dermatitis (atopic, contact, and seborrheic dermatitis), 0.29% for acne vulgaris, 0.19% for psoriasis, 0.19% for urticaria, 0.16% for viral skin diseases, 0.15% for fungal skin diseases, 0.07% for scabies, 0.06% for malignant skin melanoma, 0.05% for pyoderma, 0.04% for cellulitis, 0.03% for keratinocyte carcinoma, 0.03% for decubitus ulcer, and 0.01% for alopecia areata. All other skin and subcutaneous diseases composed 0.12% of total DALYs. Conclusions and Relevance Skin and subcutaneous diseases were the 18th leading cause of global DALYs in Global Burden of Disease 2013. Excluding mortality, skin diseases were the fourth leading cause of disability worldwide.

Forecasting peaks of seasonal influenza epidemics.

We present a framework for near real-time forecast of influenza epidemics using a simulation optimization approach. The method combines an individual-based model and a simple root finding optimization method for parameter estimation and forecasting. In this study, retrospective forecasts were generated for seasonal influenza epidemics using web-based estimates of influenza activity from Google Flu Trends for 2004-2005, 2007-2008 and 2012-2013 flu seasons. In some cases, the peak could be forecasted 5-6 weeks ahead. This study adds to existing resources for influenza forecasting and the proposed method can be used in conjunction with other approaches in an ensemble framework.