TY - JOUR AU - Quandelacy, Talia AU - Zimmer, Shanta AU - Vukotich, Chuck AU - Bieltz, Rachel AU - Grantz, Kyra AU - Galloway, David AU - Lessler, Justin AU - Zheteyeva, Yenlik AU - Uzicanin, Amra AU - Gao, Hongjiang AU - Cummings, Derek PY - 2017/05/02 Y2 - 2024/03/28 TI - Predicting virologically confirmed influenza using school absences in PA JF - Online Journal of Public Health Informatics JA - OJPHI VL - 9 IS - 1 SE - Communicable Disease Surveillance Use Cases for Human, Animal, and Zoonotic Diseases DO - 10.5210/ojphi.v9i1.7701 UR - https://ojphi.org/ojs/index.php/ojphi/article/view/7701 SP - AB - <div style="left: 78.2727px; top: 339.841px; font-size: 12.3207px; font-family: sans-serif; transform: scaleX(1.0823);" data-canvas-width="55.467823232323234">Objective</div><div style="left: 91.3182px; top: 352.844px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.02039);" data-canvas-width="331.8483883838383">To determine if all-cause and cause-specific school absences</div><div style="left: 78.2727px; top: 367.339px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.02119);" data-canvas-width="344.8553588383837">improve predictions of virologically confirmed influenza in the</div><div style="left: 78.2727px; top: 381.834px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.01567);" data-canvas-width="59.213318181818174">community.</div><div style="left: 78.2727px; top: 409.417px; font-size: 12.3207px; font-family: sans-serif; transform: scaleX(1.11884);" data-canvas-width="71.86668434343434">Introduction</div><div style="left: 91.3182px; top: 422.42px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.03316);" data-canvas-width="331.9370974747474">School-based influenza surveillance has been considered for</div><div style="left: 78.2727px; top: 436.915px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.01029);" data-canvas-width="342.5057999999999">real-time monitoring of influenza, as children 5-17 years old play an</div><div style="left: 78.2727px; top: 451.41px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.00157);" data-canvas-width="238.54120959595957">important role in community-level transmission.</div><div style="left: 78.2727px; top: 478.993px; font-size: 12.3207px; font-family: sans-serif; transform: scaleX(1.07424);" data-canvas-width="50.65042676767676">Methods</div><div style="left: 91.3182px; top: 491.996px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.969374);" data-canvas-width="328.61789898989906">The Allegheny County Department of Health provided virologically</div><div style="left: 78.2727px; top: 506.491px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.02451);" data-canvas-width="342.88527777777756">confirmed influenza data collected from all emergency departments</div><div style="left: 78.2727px; top: 520.985px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.995107);" data-canvas-width="344.124740909091">and outpatient providers in the county for 2007 and 2011-2016.</div><div style="left: 78.2727px; top: 535.48px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.00635);" data-canvas-width="344.67547651515133">All-cause school absence rates were collected from nine school</div><div style="left: 78.2727px; top: 549.975px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.972034);" data-canvas-width="343.8894154040404">districts within Allegheny County for 2010-2015. For a subset of</div><div style="left: 78.2727px; top: 564.47px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.0003);" data-canvas-width="342.16821262626263">these schools, in addition to all-cause absences, influenza-like illness</div><div style="left: 78.2727px; top: 578.965px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.996808);" data-canvas-width="344.11488434343414">(ILI)-specific absences were collected using a standard protocol:</div><div style="left: 78.2727px; top: 593.46px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.995017);" data-canvas-width="342.08689595959595">10 K-5 schools in one school district (2007-2008), nine K-12 schools</div><div style="left: 78.2727px; top: 607.955px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.984378);" data-canvas-width="341.79243106060596">in two school districts (2012-2013), and nine K-12 schools from three</div><div style="left: 78.2727px; top: 622.45px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.02456);" data-canvas-width="343.15879747474725">school districts (2015-2016). We used negative binomial regression</div><div style="left: 78.2727px; top: 636.945px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.01054);" data-canvas-width="342.45405303030304">to predict weekly county-level influenza cases in Allegheny County,</div><div style="left: 78.2727px; top: 651.44px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.01357);" data-canvas-width="342.58588459595967">Pennsylvania, during the 2010-2015 influenza seasons. We included</div><div style="left: 78.2727px; top: 665.935px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.972933);" data-canvas-width="341.46716439393913">the following covariates in candidate models: all-cause school absence</div><div style="left: 78.2727px; top: 680.43px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.00794);" data-canvas-width="342.3973777777778">rates with different lags (weekly, 1-3 week lags, assessed in separate</div><div style="left: 78.2727px; top: 694.925px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.01312);" data-canvas-width="342.63639949494944">models using all other covariates) and administrative levels (county,</div><div style="left: 78.2727px; top: 709.42px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.955599);" data-canvas-width="343.64916161616173">school type, and grade), week and month of the year (assessed in</div><div style="left: 78.2727px; top: 723.915px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.02441);" data-canvas-width="343.0589997474749">separate models), average weekly temperature, and average weekly</div><div style="left: 78.2727px; top: 738.41px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.04497);" data-canvas-width="345.5034280303028">relative humidity. Separately, for the three districts for which</div><div style="left: 78.2727px; top: 752.905px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.00208);" data-canvas-width="344.21591414141403">ILI-specific and all-cause absences were available, we predicted</div><div style="left: 78.2727px; top: 767.4px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.01268);" data-canvas-width="342.6055977272727">weekly county-level influenza cases using all-cause and ILI-specific</div><div style="left: 78.2727px; top: 781.895px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.991803);" data-canvas-width="339.31227272727244">absences with all previously stated covariates. We used several cross-</div><div style="left: 78.2727px; top: 796.39px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.97785);" data-canvas-width="341.85649873737356">validation approaches to assess models, including leave 20% of weeks</div><div style="left: 78.2727px; top: 810.884px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.00132);" data-canvas-width="271.0062727272727">out, leave 20% of schools out, and leave 52-weeks out.</div><div style="left: 78.2727px; top: 838.467px; font-size: 12.3207px; font-family: sans-serif; transform: scaleX(1.08542);" data-canvas-width="44.50239393939395">Results</div><div style="left: 91.3182px; top: 851.47px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.02135);" data-canvas-width="331.97405959595955">Overall, 2,395,020 all-cause absences were observed in nine</div><div style="left: 78.2727px; top: 865.965px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.991462);" data-canvas-width="341.8725156565656">school districts. From the subset of schools that collected ILI-specific</div><div style="left: 78.2727px; top: 880.46px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.01372);" data-canvas-width="344.7321517676764">absences, 14,078 all-cause and 2,617 ILI-related absences were</div><div style="left: 78.2727px; top: 894.955px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.04017);" data-canvas-width="343.2561310606062">reported. A total of 11,946 virologically confirmed influenza cases</div><div style="left: 78.2727px; top: 909.45px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.02539);" data-canvas-width="343.0602318181815">were reported in Allegheny County (Figure 1). Inclusion of 1-week</div><div style="left: 78.2727px; top: 923.945px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.00362);" data-canvas-width="342.30127626262635">lagged absence rates in multivariate models improved model fits and</div><div style="left: 78.2727px; top: 938.44px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.04217);" data-canvas-width="342.9961641414142">predictions of influenza cases over models using week of year and</div><div style="left: 78.2727px; top: 952.935px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.976856);" data-canvas-width="341.59037146464647">weekly average temperature (change in AIC=-4). Using grade-specific</div><div style="left: 78.2727px; top: 967.43px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.01653);" data-canvas-width="342.8594042929293">all-cause absences, absences from lower grades explained data best.</div><div style="left: 78.2727px; top: 981.925px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.02193);" data-canvas-width="344.8664474747474">For example, kindergarten absences explained 22.1% of model</div><div style="left: 78.2727px; top: 996.42px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.00088);" data-canvas-width="188.11625176767677">deviance compared to 0.43% using 12</div><div style="left: 266.37px; top: 996.746px; font-size: 7.39242px; font-family: serif; transform: scaleX(1.00023);" data-canvas-width="5.75130606060606">th</div><div style="left: 272.12px; top: 996.42px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.00072);" data-canvas-width="148.42386186868686">grade absences in validation.</div><div style="left: 78.2727px; top: 1010.91px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.01033);" data-canvas-width="342.6351674242424">Multivariate models of week-lagged kindergarten absences, week of</div><div style="left: 78.2727px; top: 1025.41px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.963526);" data-canvas-width="343.70583686868684">year, and weekly average temperature had the best fits over other</div><div style="left: 78.2727px; top: 1039.9px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.992309);" data-canvas-width="344.00030176767655">grade-specific multivariate models (change in AIC=-6 comparing</div><div style="left: 78.2727px; top: 1054.4px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.971862);" data-canvas-width="35.87789898989899">K to 12</div><div style="left: 114.156px; top: 1054.73px; font-size: 7.39242px; font-family: serif; transform: scaleX(0.995085);" data-canvas-width="5.721736363636364">th</div><div style="left: 119.88px; top: 1054.4px; font-size: 12.3207px; font-family: serif; transform: scaleX(0.983405);" data-canvas-width="300.1188714646464">grade). The utility of ILI-specific absences compared to total</div><div style="left: 443.545px; top: 338.349px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.01629);" data-canvas-width="342.7423575757576">absences is mixed, performing marginally better, adjusting for other</div><div style="left: 443.545px; top: 352.844px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.01168);" data-canvas-width="342.452820959596">covariates, in 2 years, but markedly worse in 1 year. However, these</div><div style="left: 443.545px; top: 367.339px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.00128);" data-canvas-width="266.5584974747474">results were based on a small number of observations.</div><div style="left: 443.545px; top: 394.922px; font-size: 12.3207px; font-family: sans-serif; transform: scaleX(1.10408);" data-canvas-width="73.93656313131314">Conclusions</div><div style="left: 456.591px; top: 407.925px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.02509);" data-canvas-width="329.9928898989898">Our findings suggest models including younger student absences</div><div style="left: 443.545px; top: 422.42px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.01912);" data-canvas-width="342.7300368686869">improve predictions of virologically confirmed influenza. We found</div><div style="left: 443.545px; top: 436.915px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.01425);" data-canvas-width="344.57444671717167">ILI-specific absences performed similarly to all-cause absences;</div><div style="left: 443.545px; top: 451.41px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.04895);" data-canvas-width="345.47016212121207">however, more observations are needed to assess the relative</div><div style="left: 443.545px; top: 465.905px; font-size: 12.3207px; font-family: serif; transform: scaleX(1.00126);" data-canvas-width="175.52079292929298">performances of these two datasets.</div> ER -