Optimizing Provider Recruitment for Influenza Surveillance Networks

Citation: Scarpino SV, Dimitrov NB, Meyers LA ( Optimizing Provider Recruitment for Influenza Surveillance Networks Samuel V. Scarpino 0 Nedialko B. Dimitrov 0 Lauren Ancel Meyers 0 Mark M. Tanaka, University of New South Wales, Australia 0 1 The University of Texas at Austin, Section of Integrative Biology, Austin, Texas, United States of America, 2 Naval Postgraduate School, Operations Research Department, Monterey, California, United States of America, 3 The Santa Fe Institute , Santa Fe, New Mexico , United States of America The increasingly complex and rapid transmission dynamics of many infectious diseases necessitates the use of new, more advanced methods for surveillance, early detection, and decision-making. Here, we demonstrate that a new method for optimizing surveillance networks can improve the quality of epidemiological information produced by typical providerbased networks. Using past surveillance and Internet search data, it determines the precise locations where providers should be enrolled. When applied to redesigning the provider-based, influenza-like-illness surveillance network (ILINet) for the state of Texas, the method identifies networks that are expected to significantly outperform the existing network with far fewer providers. This optimized network avoids informational redundancies and is thereby more effective than networks designed by conventional methods and a recently published algorithm based on maximizing population coverage. We show further that Google Flu Trends data, when incorporated into a network as a virtual provider, can enhance but not replace traditional surveillance methods. - Since the Spanish Flu Pandemic of 1918{1919, the global public health community has made great strides towards the effective surveillance of infectious diseases. However, modern travel patterns, heterogeneity in human population densities, proximity to wildlife populations, and variable immunity interact to drive increasingly complex patterns of disease transmission and emergence. As a result, there is an increasing need for effective, evidence-based surveillance, early detection, and decision-making methods [13]. This need was clearly articulated in 2009 by a directive from the Department of Homeland Security and the Centers for Disease Control and Prevention to develop a nationwide, real-time public health surveillance network [4,5]. The U.S. Outpatient Influenza-Like Illness Surveillance Network (ILINet) gathers data from thousands of healthcare providers across all fifty states. Throughout influenza season (CDC mandating reporting during weeks 40{20, which is approximately October through mid-May), participating providers are asked to report weekly the number of cases of influenza-like illness treated and total number of patients seen, by age group. Cases qualify as ILI if they manifest fever in excess of 1000F along with a cough and/or a sore throat, without another known cause. Although the CDC receives reports of approximately 16 million patient visits per year, many of the reports may use a loose application of the ILI case definition and/or may simply be inaccurate. The data are used in conjunction with other sources of laboratory, hospitalization and mortality data to monitor regional and national influenza activity and associated mortality. Similar national surveillance networks are in place in 11 EU countries and elsewhere around the globe [69]. Each US state is responsible for recruiting and managing ILINet providers. The CDC advises states to recruit one regularly reporting sentinel provider per 250,000 residents, with a state-wide minimum of 10 sentinel providers. Since 2003, the Texas Department of State Health Services (DSHS) has enrolled a total of 300 volunteer providers. Participating providers regularly drop out of the network; Texas DSHS aims to maintain approximately 200 active participants through year-round recruitment of providers in heavily populated areas (cities with populations of at least 100,000). DSHS also permits other (non-targeted) providers of family medicine, internal medicine, pediatrics, university student health services, emergency medicine, infectious disease, OB/GYN and urgent care to participate in the network. During the 2009{2010 influenza season, the Texas ILINet included 205 providers with approximately 50% reporting most weeks of the influenza season. A number of statistical studies have demonstrated that ILI surveillance data is adequate for characterizing past influenza epidemics, monitoring populations for abnormal influenza activity, and forecasting the onsets and peaks of local influenza epidemics [1016]. However, the surveillance networks are often limited by non-representative samples [17], inaccurate and variable reporting [1214], and low reporting rates [6]. Some of these studies have yielded specific recommendations for improving the performance of the surveillance network, for example, inclusion of particular categories of hospitals in China [12], preference for general practitioners over pediatricians in Paris, France [14], and a Public health agencies use surveillance systems to detect and monitor chronic and infectious diseases. These systems often rely on data sources that are chosen based on loose guidelines or out of convenience. In this paper, we introduce a new, data-driven method for designing and improving surveillance systems. Our approach is a geographic optimization of data sources designed to achieve specific surveillance goals. We tested our method by re-designing Texas provider-based influenza surveillance system (ILINet). The resulting networks better predicted influenza associated hospitalizations and contained fewer providers than the existing ILINet. Furthermore, our study demonstrates that the integration of Internet source data, like Google Flu Trends, into surveillance systems can enhance traditional, providerbased networks. general guideline to target practices with high reporting rates and high numbers of patient visits (per capita) [6]. Polgreen et al. (2009) recently described a computational method for selecting ILINet providers so as to maximize coverage, that is, the number of people living within a specified distance of a provider [17]. They applied the approach to optimizing the placement of the 22 providers in the Iowa ILINet. While their algorithm ensures maximum coverage, it is not clear that maximum coverage is, in general, the most appropriate criterion for building a statistically informative ILINet. In 2008, Google.org launched Google Flu Trends, a website that translates the daily number of Googles search terms associated with signs, symptoms, and treatment for acute respiratory infections into an estimate of the number of ILI patients per 100,000 people. It was shown that Google Flu Trends reliably estimates national influenza activity in the US [18], the state of Utah [18], and in some European countries [19], but it provided imperfect data re (...truncated)