@article{Aronis_Millett_Wagner_Tsui_Ye_Ferraro_Haug_Cooper_2017, title={Detecting Overlapping Outbreaks of Influenza}, volume={9}, url={https://ojphi.org/ojs/index.php/ojphi/article/view/7592}, DOI={10.5210/ojphi.v9i1.7592}, abstractNote={ntroduction<div style="left: 105px; top: 329.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.04226);" data-canvas-width="379.40033333333326">Influenza is a contagious disease that causes epidemics in many</div><div style="left: 90px; top: 345.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.971632);" data-canvas-width="394.91991666666667">parts of the world. The World Health Organization estimates that</div><div style="left: 90px; top: 362.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.02669);" data-canvas-width="394.3362500000001">influenza causes three to five million severe illnesses each year and</div><div style="left: 90px; top: 379.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.00595);" data-canvas-width="393.8304999999998">250,000-500,000 deaths [1]. Predicting and characterizing outbreaks</div><div style="left: 90px; top: 395.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.965183);" data-canvas-width="394.8080000000001">of influenza is an important public health problem and significant</div><div style="left: 90px; top: 412.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.979035);" data-canvas-width="395.22024999999985">progress has been made in predicting single outbreaks. However,</div><div style="left: 90px; top: 429.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.06456);" data-canvas-width="398.18249999999983">multiple temporally overlapping outbreaks are also common.</div><div style="left: 90px; top: 445.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.02063);" data-canvas-width="394.23708333333343">These may be caused by different subtypes or outbreaks in multiple</div><div style="left: 90px; top: 462.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.02796);" data-canvas-width="224.3858333333333">demographic groups. We describe our</div><div style="left: 314.305px; top: 462.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.01864);" data-canvas-width="170.1983333333333">Multiple Outbreak Detection</div><div style="left: 90px; top: 479.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.978673);" data-canvas-width="40.794333333333334">System</div><div style="left: 130.817px; top: 479.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.02477);" data-canvas-width="355.88508333333317">(MODS) and its performance on two actual outbreaks.</div><div style="left: 90px; top: 495.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.968229);" data-canvas-width="390.16416666666646">This work extends previous work by our group [2,3,4] by using model-</div><div style="left: 90px; top: 512.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.997086);" data-canvas-width="393.23408333333344">averaging and a new method to estimate non-influenza influenza-like</div><div style="left: 90px; top: 529.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.997503);" data-canvas-width="393.39841666666666">illness (NI-ILI). We also apply MODS to a real dataset with a double</div><div style="left: 90px; top: 545.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.00083);" data-canvas-width="53.11083333333333">outbreak.</div><div style="left: 90px; top: 577.425px; font-size: 14.1667px; font-family: sans-serif; transform: scaleX(1.07287);" data-canvas-width="58.23916666666666">Methods</div><div style="left: 105px; top: 592.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.0196);" data-canvas-width="379.25583333333327">MODS is part of a framework for disease surveillance developed</div><div style="left: 90px; top: 609.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.974191);" data-canvas-width="392.6801666666667">by our group. In this framework, a natural language processing system</div><div style="left: 90px; top: 625.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.0112);" data-canvas-width="394.09825000000006">extracts symptoms from emergency department patient-care reports.</div><div style="left: 90px; top: 642.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.02014);" data-canvas-width="394.21583333333297">These features are combined with laboratory results and passed to a</div><div style="left: 90px; top: 659.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.03639);" data-canvas-width="394.5530000000001">case detection system that infers a probability distribution over the</div><div style="left: 90px; top: 675.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.972066);" data-canvas-width="395.19333333333316">diseases each patient may have. These diseases include influenza,</div><div style="left: 90px; top: 692.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.983445);" data-canvas-width="395.60699999999986">NI-ILI, and other (appendicitis, trauma, etc.). This distribution is</div><div style="left: 90px; top: 709.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.01305);" data-canvas-width="393.9749999999999">expressed in terms of the likelihoods of the patients’ data. These are</div><div style="left: 90px; top: 725.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.994455);" data-canvas-width="393.3233333333334">given to MODS which searches a space of multiple outbreak models,</div><div style="left: 90px; top: 742.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.02488);" data-canvas-width="394.4424999999996">computes the likelihood of each model, and calculates the expected</div><div style="left: 90px; top: 759.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.03641);" data-canvas-width="394.47508333333315">number of influenza cases day-by-day. This work differs from past</div><div style="left: 90px; top: 775.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.993312);" data-canvas-width="396.14958333333317">work in three important ways. First, we address the problem of</div><div style="left: 90px; top: 792.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.984035);" data-canvas-width="393.03999999999974">detecting and characterizing multiple, overlapping outbreaks. Second,</div><div style="left: 90px; top: 809.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.01748);" data-canvas-width="394.10249999999996">we do not rely on simple counts, but use likelihoods given evidence</div><div style="left: 90px; top: 825.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.03682);" data-canvas-width="394.23283333333336">in the free-text portion of patient-care reports as well as laboratory</div><div style="left: 90px; top: 842.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.977935);" data-canvas-width="389.9516666666664">findings. Third, we explicitly account for non-influenza influenza-</div><div style="left: 90px; top: 859.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.99518);" data-canvas-width="393.2454166666667">like illnesses. This is important because some forms of influenza-like</div><div style="left: 90px; top: 875.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.988156);" data-canvas-width="393.0555833333334">illness (such as respiratory syncytial virus) are contagious and exhibit</div><div style="left: 90px; top: 892.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.03851);" data-canvas-width="394.6691666666666">outbreak activity. This research was approved by the University of</div><div style="left: 90px; top: 909.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.0017);" data-canvas-width="267.1691666666667">Pittsburgh and Intermountain Healthcare IRBs.</div><div style="left: 90px; top: 940.758px; font-size: 14.1667px; font-family: sans-serif; transform: scaleX(1.08488);" data-canvas-width="51.17">Results</div><div style="left: 105px; top: 955.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.996423);" data-canvas-width="380.89916666666653">We conducted a set of experiments with simulated outbreaks.</div><div style="left: 90px; top: 972.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.03333);" data-canvas-width="394.4906666666665">MODS is able to detect a single outbreak six to eight weeks before</div><div style="left: 90px; top: 989.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.98929);" data-canvas-width="393.16041666666683">the peak. It is also able to recognize a second outbreak approximately</div><div style="left: 90px; top: 1005.71px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.08039);" data-canvas-width="398.6103333333333">halfway between peaks for simulated double outbreaks. We</div><div style="left: 90px; top: 1022.38px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.02748);" data-canvas-width="396.7091666666666">conducted experiments using real outbreaks and compared our</div><div style="left: 90px; top: 1039.04px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.01333);" data-canvas-width="394.0656666666667">results to thermometer sales [5]. Using data from Allegheny County</div><div style="left: 90px; top: 1055.71px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.00102);" data-canvas-width="395.7033333333334">Pennsylvania for the 2009-2010 influenza season, on September</div><div style="left: 90px; top: 1072.38px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.991683);" data-canvas-width="395.91300000000024">1 MODS predicted an outbreak with a peak on October 5. The</div><div style="left: 90px; top: 1089.04px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.984709);" data-canvas-width="392.9975">thermometer peak was October 21. The figure “Prediction on October</div><div style="left: 90px; top: 1105.71px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.00992);" data-canvas-width="393.9183333333333">1 for Allegheny County” compares MODS’ prediction on October 1</div><div style="left: 90px; top: 1122.38px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.04347);" data-canvas-width="394.3277499999999">to thermometer sales. Using data from Salt Lake City Utah for the</div><div style="left: 90px; top: 1139.04px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.973638);" data-canvas-width="395.0374999999999">2010-2011 influenza season, on November 1 MODS predicted an</div><div style="left: 90px; top: 1155.71px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.01413);" data-canvas-width="393.97075">outbreak with peak on December 7. The first thermometer peak was</div><div style="left: 90px; top: 1172.38px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.960601);" data-canvas-width="394.7909999999998">December 29. On January 20 MODS predicted a second outbreak</div><div style="left: 90px; top: 1189.04px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.02121);" data-canvas-width="394.2724999999998">with peak on February 9. The second thermometer peak was March</div><div style="left: 510px; top: 312.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.990985);" data-canvas-width="393.17316666666676">5. The figure “Prediction on January 20 for Salt Lake City” compares</div><div style="left: 510px; top: 329.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.00054);" data-canvas-width="314.76916666666654">MODS’ prediction on January 20 to thermometer sales.</div><div style="left: 510px; top: 360.758px; font-size: 14.1667px; font-family: sans-serif; transform: scaleX(1.10336);" data-canvas-width="85.01416666666667">Conclusions</div><div style="left: 525px; top: 375.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.977373);" data-canvas-width="103.48750000000001">We have built a</div><div style="left: 628.532px; top: 375.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.993554);" data-canvas-width="218.00233333333333">Multiple Outbreak Detection System</div><div style="left: 846.664px; top: 375.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.949812);" data-canvas-width="59.04666666666667">that can</div><div style="left: 510px; top: 392.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(0.997328);" data-canvas-width="393.2964166666667">detect and characterize overlapping outbreaks of influenza. Although</div><div style="left: 510px; top: 409.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.03547);" data-canvas-width="394.5813333333334">the system currently predicts outbreaks of influenza, it is built on a</div><div style="left: 510px; top: 425.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.02043);" data-canvas-width="394.3263333333333">general Bayesian framework that can be extended to other diseases.</div><div style="left: 510px; top: 442.376px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.01058);" data-canvas-width="396.19916666666666">Future work includes incorporating multiple forms of evidence,</div><div style="left: 510px; top: 459.043px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.01613);" data-canvas-width="394.30791666666664">modeling other known contagious diseases, and detecting outbreaks</div><div style="left: 510px; top: 475.709px; font-size: 14.1667px; font-family: serif; transform: scaleX(1.00067);" data-canvas-width="212.075">of new previously unknown diseases.</div><div style="left: 510px; top: 739.394px; font-size: 12.5px; font-family: serif; transform: scaleX(1.00063);" data-canvas-width="289.9">Prediction on October 1 for Allegheny County 2009-2010</div><div style="left: 510px; top: 1009.06px; font-size: 12.5px; font-family: serif; transform: scaleX(1.0024);" data-canvas-width="274.975">Prediction on January 20 for Salt Lake City 2010-2011</div>}, number={1}, journal={Online Journal of Public Health Informatics}, author={Aronis, John and Millett, Nicholas and Wagner, Michael and Tsui, Fuchiang and Ye, Ye and Ferraro, Jeffrey and Haug, Peter and Cooper, Gregory}, year={2017}, month={May} }