Donald R. Olson

Reassessing Google Flu Trends Data for Detection of Seasonal and Pandemic Influenza: A Comparative Epidemiological Study at Three Geographic Scales

The goal of influenza-like illness (ILI) surveillance is to determine the timing, location and magnitude of outbreaks by monitoring the frequency and progression of clinical case incidence. Advances in computational and information technology have allowed for automated collection of higher volumes of electronic data and more timely analyses than previously possible. Novel surveillance systems, including those based on internet search query data like Google Flu Trends (GFT), are being used as surrogates for clinically-based reporting of influenza-like-illness (ILI). We investigated the reliability of GFT during the last decade (2003 to 2013), and compared weekly public health surveillance with search query data to characterize the timing and intensity of seasonal and pandemic influenza at the national (United States), regional (Mid-Atlantic) and local (New York City) levels. We identified substantial flaws in the original and updated GFT models at all three geographic scales, including completely missing the first wave of the 2009 influenza A/H1N1 pandemic, and greatly overestimating the intensity of the A/H3N2 epidemic during the 2012/2013 season. These results were obtained for both the original (2008) and the updated (2009) GFT algorithms. The performance of both models was problematic, perhaps because of changes in internet search behavior and differences in the seasonality, geographical heterogeneity and age-distribution of the epidemics between the periods of GFT model-fitting and prospective use. We conclude that GFT data may not provide reliable surveillance for seasonal or pandemic influenza and should be interpreted with caution until the algorithm can be improved and evaluated. Current internet search query data are no substitute for timely local clinical and laboratory surveillance, or national surveillance based on local data collection. New generation surveillance systems such as GFT should incorporate the use of near-real time electronic health data and computational methods for continued model-fitting and ongoing evaluation and improvement.

Monitoring the Impact of Influenza by Age: Emergency Department Fever and Respiratory Complaint Surveillance in New York City

Background The importance of understanding age when estimating the impact of influenza on hospitalizations and deaths has been well described, yet existing surveillance systems have not made adequate use of age-specific data. Monitoring influenza-related morbidity using electronic health data may provide timely and detailed insight into the age-specific course, impact and epidemiology of seasonal drift and reassortment epidemic viruses. The purpose of this study was to evaluate the use of emergency department (ED) chief complaint data for measuring influenza-attributable morbidity by age and by predominant circulating virus. Methods and Findings We analyzed electronically reported ED fever and respiratory chief complaint and viral surveillance data in New York City (NYC) during the 2001–2002 through 2005–2006 influenza seasons, and inferred dominant circulating viruses from national surveillance reports. We estimated influenza-attributable impact as observed visits in excess of a model-predicted baseline during influenza periods, and epidemic timing by threshold and cross correlation. We found excess fever and respiratory ED visits occurred predominantly among school-aged children (8.5 excess ED visits per 1,000 children aged 5–17 y) with little or no impact on adults during the early-2002 B/Victoria-lineage epidemic; increased fever and respiratory ED visits among children younger than 5 y during respiratory syncytial virus-predominant periods preceding epidemic influenza; and excess ED visits across all ages during the 2003–2004 (9.2 excess visits per 1,000 population) and 2004–2005 (5.2 excess visits per 1,000 population) A/H3N2 Fujian-lineage epidemics, with the relative impact shifted within and between seasons from younger to older ages. During each influenza epidemic period in the study, ED visits were increased among school-aged children, and each epidemic peaked among school-aged children before other impacted age groups. Conclusions Influenza-related morbidity in NYC was highly age- and strain-specific. The impact of reemerging B/Victoria-lineage influenza was focused primarily on school-aged children born since the virus was last widespread in the US, while epidemic A/Fujian-lineage influenza affected all age groups, consistent with a novel antigenic variant. The correspondence between predominant circulating viruses and excess ED visits, hospitalizations, and deaths shows that excess fever and respiratory ED visits provide a reliable surrogate measure of incident influenza-attributable morbidity. The highly age-specific impact of influenza by subtype and strain suggests that greater age detail be incorporated into ongoing surveillance. Influenza morbidity surveillance using electronic data currently available in many jurisdictions can provide timely and representative information about the age-specific epidemiology of circulating influenza viruses.

Prioritization of Influenza Pandemic Vaccination to Minimize Years of Life Lost

BACKGROUND How to allocate limited vaccine supplies in the event of an influenza pandemic is currently under debate. Conventional vaccination strategies focus on those at highest risk for severe outcomes, including seniors, but do not consider (1) the signature pandemic pattern in which mortality risk is shifted to younger ages, (2) likely reduced vaccine response in seniors, and (3) differences in remaining years of life with age. METHODS We integrated these factors to project the age-specific years of life lost (YLL) and saved in a future pandemic, on the basis of mortality patterns from 3 historical pandemics, age-specific vaccine efficacy, and the 2000 US population structure. RESULTS For a 1918-like scenario, the absolute mortality risk is highest in people <45 years old; in contrast, seniors (those >or=65 years old) have the highest mortality risk in the 1957 and 1968 scenarios. The greatest YLL savings would be achieved by targeting different age groups in each scenario; people <45 years old in the 1918 scenario, people 45-64 years old in the 1968 scenario, and people >45 years old in the 1957 scenario. CONCLUSIONS Our findings shift the focus of pandemic vaccination strategies onto younger populations and illustrate the need for real-time surveillance of mortality patterns in a future pandemic. Flexible setting of vaccination priority is essential to minimize mortality.

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.

Demonstrating the Use of High-Volume Electronic Medical Claims Data to Monitor Local and Regional Influenza Activity in the US

Introduction Fine-grained influenza surveillance data are lacking in the US, hampering our ability to monitor disease spread at a local scale. Here we evaluate the performances of high-volume electronic medical claims data to assess local and regional influenza activity. Material and Methods We used electronic medical claims data compiled by IMS Health in 480 US locations to create weekly regional influenza-like-illness (ILI) time series during 2003–2010. IMS Health captured 62% of US outpatient visits in 2009. We studied the performances of IMS-ILI indicators against reference influenza surveillance datasets, including CDC-ILI outpatient and laboratory-confirmed influenza data. We estimated correlation in weekly incidences, peak timing and seasonal intensity across datasets, stratified by 10 regions and four age groups (<5, 5–29, 30–59, and 60+ years). To test IMS-Health performances at the city level, we compared IMS-ILI indicators to syndromic surveillance data for New York City. We also used control data on laboratory-confirmed Respiratory Syncytial Virus (RSV) activity to test the specificity of IMS-ILI for influenza surveillance. Results Regional IMS-ILI indicators were highly synchronous with CDCs reference influenza surveillance data (Pearson correlation coefficients rho≥0.89; range across regions, 0.80–0.97, P<0.001). Seasonal intensity estimates were weakly correlated across datasets in all age data (rho≤0.52), moderately correlated among adults (rho≥0.64) and uncorrelated among school-age children. IMS-ILI indicators were more correlated with reference influenza data than control RSV indicators (rho = 0.93 with influenza v. rho = 0.33 with RSV, P<0.05). City-level IMS-ILI indicators were highly consistent with reference syndromic data (rho≥0.86). Conclusion Medical claims-based ILI indicators accurately capture weekly fluctuations in influenza activity in all US regions during inter-pandemic and pandemic seasons, and can be broken down by age groups and fine geographical areas. Medical claims data provide more reliable and fine-grained indicators of influenza activity than other high-volume electronic algorithms and should be used to augment existing influenza surveillance systems.

Forecasting Influenza Outbreaks in Boroughs and Neighborhoods of New York City

The ideal spatial scale, or granularity, at which infectious disease incidence should be monitored and forecast has been little explored. By identifying the optimal granularity for a given disease and host population, and matching surveillance and prediction efforts to this scale, response to emergent and recurrent outbreaks can be improved. Here we explore how granularity and representation of spatial structure affect influenza forecast accuracy within New York City. We develop network models at the borough and neighborhood levels, and use them in conjunction with surveillance data and a data assimilation method to forecast influenza activity. These forecasts are compared to an alternate system that predicts influenza for each borough or neighborhood in isolation. At the borough scale, influenza epidemics are highly synchronous despite substantial differences in intensity, and inclusion of network connectivity among boroughs generally improves forecast accuracy. At the neighborhood scale, we observe much greater spatial heterogeneity among influenza outbreaks including substantial differences in local outbreak timing and structure; however, inclusion of the network model structure generally degrades forecast accuracy. One notable exception is that local outbreak onset, particularly when signal is modest, is better predicted with the network model. These findings suggest that observation and forecast at sub-municipal scales within New York City provides richer, more discriminant information on influenza incidence, particularly at the neighborhood scale where greater heterogeneity exists, and that the spatial spread of influenza among localities can be forecast.

Applying a New Model for Sharing Population Health Data to National Syndromic Influenza Surveillance: DiSTRIBuTE Project Proof of Concept, 2006 to 2009.

The Distributed Surveillance Taskforce for Real-time Influenza Burden Tracking and Evaluation (DiSTRIBuTE) project began as a pilot effort initiated by the International Society for Disease Surveillance (ISDS) in autumn 2006 to create a collaborative electronic emergency department (ED) syndromic influenza-like illness (ILI) surveillance network based on existing state and local systems and expertise. DiSTRIBuTE brought together health departments that were interested in: 1) sharing aggregate level data; 2) maintaining jurisdictional control; 3) minimizing barriers to participation; and 4) leveraging the flexibility of local systems to create a dynamic and collaborative surveillance network. This approach was in contrast to the prevailing paradigm for surveillance where record level information was collected, stored and analyzed centrally. The DiSTRIBuTE project was created with a distributed design, where individual level data remained local and only summarized, stratified counts were reported centrally, thus minimizing privacy risks. The project was responsive to federal mandates to improve integration of federal, state, and local biosurveillance capabilities. During the proof of concept phase, 2006 to 2009, ten jurisdictions from across North America sent ISDS on a daily to weekly basis year-round, aggregated data by day, stratified by local ILI syndrome, age-group and region. During this period, data from participating U.S. state or local health departments captured over 13% of all ED visits nationwide. The initiative focused on state and local health department trust, expertise, and control. Morbidity trends observed in DiSTRIBuTE were highly correlated with other influenza surveillance measures. With the emergence of novel A/H1N1 influenza in the spring of 2009, the project was used to support information sharing and ad hoc querying at the state and local level. In the fall of 2009, through a broadly collaborative effort, the project was expanded to enhance electronic ED surveillance nationwide.

Estimating the hospitalization burden associated with influenza and respiratory syncytial virus in New York City, 2003–2011

Hospitalization burden associated with influenza and respiratory syncytial virus (RSV) is uncertain due to ambiguity in the inference methodologies employed for its estimation.

Mortality and morbidity burden associated with A/H1N1pdm influenza virus: Who is likely to be infected, experience clinical symptoms, or die from the H1N1pdm 2009 pandemic virus ?

Here we use lessons from past influenza pandemics and recent information about the H1N1pdm pandemic to discuss variations in H1N1pdm disease burden with age, underlying risk factors, and geography.

Perceived usefulness of a distributed community-based syndromic surveillance system: a pilot qualitative evaluation study

BackgroundWe conducted a pilot utility evaluation and information needs assessment of the Distribute Project at the 2010 Washington State Public Health Association (WSPHA) Joint Conference. Distribute is a distributed community-based syndromic surveillance system and network for detection of influenza-like illness (ILI). Using qualitative methods, we assessed the perceived usefulness of the Distribute system and explored areas for improvement. Nine state and local public health professionals participated in a focus group (n = 6) and in semi-structured interviews (n = 3). Field notes were taken, summarized and analyzed.FindingsSeveral emergent themes that contribute to the perceived usefulness of system data and the Distribute system were identified: 1) Standardization: a common ILI syndrome definition; 2) Regional Comparability: views that support county-by-county comparisons of syndromic surveillance data; 3) Completeness: complete data for all expected data at a given time; 4) Coverage: data coverage of all jurisdictions in WA state; 5) Context: metadata incorporated into the views to provide context for graphed data; 6) Trusted Data: verification that information is valid and timely; and 7) Customization: the ability to customize views as necessary. As a result of the focus group, a new county level health jurisdiction expressed interest in contributing data to the Distribute system.ConclusionThe resulting themes from this study can be used to guide future information design efforts for the Distribute system and other syndromic surveillance systems. In addition, this study demonstrates the benefits of conducting a low cost, qualitative evaluation at a professional conference.