identifying and ranking common covid-19 …...2020/06/10  · twitter platform enables obtaining...

13
Original Paper Eisa Alanazi, BSc MSc PhD; Center of Innovation and Development in Artificial Intelligence Umm Al-Qura University Saudi Arabia Abdulaziz Alashaikh BEng MSc PhD; Department of Computer Engineering and Network University of Jeddah Saudi Arabia Sarah Alqurashi BSc; Center of Innovation and Development in Artificial Intelligence Umm Al-Qura University Saudi Arabia Aued Alanazi BSc; Center of Innovation and Development in Artificial Intelligence Umm Al-Qura University Saudi Arabia Corresponding Author: Eisa Alanazi, BSc MSc PhD Center of Innovation and Development in Artificial Intelligence Umm Al-Qura University Makkah, 21421 Saudi Arabia Email: [email protected] Identifying and Ranking Common COVID-19 Symptoms from Arabic Twitter Abstract Background: Massive amount of covid-19 related data is generated everyday by Twitter users. Self-reports of covid-19 symptoms on Twitter can reveal a great deal about the disease and its prevalence in the community. In particular, self-reports can be used as a valuable resource to learn more about the common symptoms and whether their order of appearance differs among different groups in the community. With sufficient available data, this has the potential of developing a covid-19 risk- assessment system that is tailored toward specific group of people. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 12, 2020. . https://doi.org/10.1101/2020.06.10.20127225 doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.

Upload: others

Post on 31-Jul-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Identifying and Ranking Common COVID-19 …...2020/06/10  · Twitter platform enables obtaining multiple features (such as age, sex, geo-location, … etc.) along with informative

Original Paper EisaAlanazi,BScMScPhD;CenterofInnovationandDevelopmentinArtificialIntelligenceUmmAl-QuraUniversitySaudiArabiaAbdulazizAlashaikhBEngMScPhD;DepartmentofComputerEngineeringandNetworkUniversityofJeddahSaudiArabiaSarahAlqurashiBSc;CenterofInnovationandDevelopmentinArtificialIntelligenceUmmAl-QuraUniversitySaudiArabiaAuedAlanaziBSc;CenterofInnovationandDevelopmentinArtificialIntelligenceUmmAl-QuraUniversitySaudiArabiaCorrespondingAuthor:EisaAlanazi,BScMScPhDCenterofInnovationandDevelopmentinArtificialIntelligenceUmmAl-QuraUniversityMakkah,21421SaudiArabiaEmail:[email protected]

Identifying and Ranking Common COVID-19 Symptoms from Arabic Twitter

Abstract Background:Massiveamountofcovid-19relateddataisgeneratedeverydaybyTwitterusers.Self-reportsofcovid-19symptomsonTwittercanrevealagreatdealaboutthediseaseanditsprevalenceinthecommunity.Inparticular,self-reportscanbeusedasavaluableresourcetolearnmoreaboutthecommonsymptomsandwhethertheirorderofappearancediffersamongdifferentgroupsinthecommunity.Withsufficientavailabledata,thishasthepotentialofdevelopingacovid-19risk-assessmentsystemthatistailoredtowardspecificgroupofpeople.

. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 12, 2020. .https://doi.org/10.1101/2020.06.10.20127225doi: medRxiv preprint

NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.

Page 2: Identifying and Ranking Common COVID-19 …...2020/06/10  · Twitter platform enables obtaining multiple features (such as age, sex, geo-location, … etc.) along with informative

Objective:Theaimofthisstudyistoidentifythemostcommonsymptomsreportedbycovid-19patientsintheArabiclanguageandorderthesymptomsappearancebasedonthecollecteddata.Methods:WesearchtheArabiccontentofTwitterforpersonalreportsofcovid-19symptomsfromMarch1sttoMay27th,2020.Weidentify463Arabicuserswhotweetedtestingpositiveforcovid-19andextractthesymptomstheypubliclyassociatewithcovid-19.Furthermore,weaskthemdirectlythroughpersonalmessagestooptinandranktheappearanceofthefirstthreesymptomstheyexperiencedrightbefore(orafter)diagnosedwithcovid-19.Finally,wetracktheirTwittertimelinetoidentifyadditionalsymptomsthatwerementionedwithin±5daysfromthedayoftweetinghavingcovid-19.Insummary,alistof270covid-19reportswerecollectedandsymptomswere(atleastpartially)rankedfromearlytolate.Results:Thecollectedreportscontainedroughly900symptomsoriginatedfrom74%(n=201)maleand26%(n=69)femaleTwitterusers.Themajority(82%)ofthetrackeduserswerelivinginSaudiArabia(46%)andKuwait(36%).Furthermore,13%(n=36)ofthecollectedreportswereasymptomatic.Outoftheuserswithsymptoms(n=234),66%(n=180)providedachronologicalorderofappearanceforatleastthreesymptoms.Fever59%(n=139),Headache43%(n=101),andAnosmia39%(n=91)werefoundtobethetopthreesymptomsmentionedbythereports.Theycountalsoforthetop-3commonfirstsymptomsinawaythat28%(n=65)saidtheircovidjourneystartedwithaFever,15%(n=34)withaHeadacheand12%(n=28)withAnosmia.OutoftheSaudisymptomaticreportedcases(n=110),themostcommonthreesymptomswereFever59%(n=65),Anosmia42%(n=46),andHeadache38%(n=42).

Conclusions:ThisstudydemonstratesthatTwitterisavaluableresourcetoanalyzeandidentifyCOVID-19earlysymptomswithintheArabiccontentofTwitter.Italsosuggeststhepossibilityofdevelopingareal-timecovid-19riskestimatorbasedontheusers’tweets.Keywords:Socialnetworksanalysis;Twitter;datamining;covid-19;earlysymptoms;ranking;Arabic;

Introduction TheongoingvigorousCOVID-19outbreakhasshownagreatimpactonhumanhealthandwell-beingandradicallyenforcedarigorouschangeinsocietieslifestyle

. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 12, 2020. .https://doi.org/10.1101/2020.06.10.20127225doi: medRxiv preprint

Page 3: Identifying and Ranking Common COVID-19 …...2020/06/10  · Twitter platform enables obtaining multiple features (such as age, sex, geo-location, … etc.) along with informative

underminingtheirprosperity.Alongwiththiscatastrophe,wehavewitnessedagreateffortfromdiverseresearchcommunitiestostudythisdiseaseinallitsaspects.Inrecentyears,socialnetworkshavebecomeanunignorablesourceofinformationwhereusersexposeandshareideas,opinions,thoughts,andexperiencesonamultitudeoftopics.Severalresearcheshaveutilizedtheabundanceofinformationofferedbysocialplatformstoconductnon-clinicalmedicalresearch.Forexample,Twitterhasbeenthesourcefordataformanyhealthandmedicalstudies;suchassurveillanceandmonitoringofFluandCancertimelineanddistributionacrosstheUSAusingTwitter[1],analyzingthespreadofinfluenzaintheUAEbasedongeo-taggedArabicTweets[2],surveillanceandmonitoringofInfluenzaintheUAEbasedonArabicandEnglishtweets[3],identifyingsymptomsanddiseaseinSaudiArabiausingTwitter[4],andmostrecentlyonanalyzingCOVID-19symptomsonTwitter[5]andanalyzingthechronologicalandgeographicaldistributionofCOVID-19infectedtweetersintheUSA[6].Twitterplatformenablesobtainingmultiplefeatures(suchasage,sex,geo-location,…etc.)alongwithinformativemessagesthatusingappropriatedataminingandanalysistechniquescanpotentiallyresultinusefulinsightsaboutaspecifichealthcondition[7].Extractingcommonsymptomsassociatedwithadiseasefrompubliclyavailabledatahasthepotentialtocontrolthespreadofthediseaseandidentifyusersathighrisk.Italsogivesnewinsightsthatcallforearlyinterventionandcontrol.Forexample,Figure1highlightsatweet(fromSaudiArabia)mentioningexplicitlythelossofsmellandtasteasonedistinctivesymptomofcovid-19.Thenational-wideself-testingquestionnaireAppwasupdatedalmostcoupleofweeksafterthetweettoincludethelosssmellandtasteasonepotentialsignforcovid-19[8].Trackingcovid-19symptomsinreal-timefrompublicdataonTwittercouldhaveshortenthegap.Figure1:Acovid-19survivortweetsabouthowthelossofsmellandtastewastheonlycommonsignamongallofherfamilymembers.

. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 12, 2020. .https://doi.org/10.1101/2020.06.10.20127225doi: medRxiv preprint

Page 4: Identifying and Ranking Common COVID-19 …...2020/06/10  · Twitter platform enables obtaining multiple features (such as age, sex, geo-location, … etc.) along with informative

Inthispaper,westudyCOVID-19symptomsasreportedbyArabictweeters.Initially,weshuffledArabictweetsandsearchingfortweetswithCOVID-19symptomsandalsocollectedtweetsforuserswhoreportedthemselvesinfectedthroughclinicaltest.Inaddition,weaskeduserswhohavebeenmarkedinfectedaboutthefirstthreesymptomstheyexperiencedviaavoluntarysurveytemplatesentoverprivatemessage.

Methods OurmethodfordatacollectionisoutlinedinFigure1.WesearchTwitterforpersonalreportsofcovid-19fromMarch1st,2020toMay27,2020usingtwoArabickeywords يتباصإ and يصیخشت whichtranslatesroughlyto“Ihavebeendiagnosed”.Suckkeywordsarelikelytofilteroutreportsthatwerenotassociatedwithaformaltestresult.Aninitiallistof463userswerecollectedandtwoindependentfreelancerswereaskedtofurtherreadusers’timelineandextractsymptomsthatwereexplicitlymentionedtoberelatedtocovidandtheirorderofappearanceifmentioned.Additionalinformationsuchasusergender,dateofinfection,andthecountryofresidencewerealsocollected.Weassumethedateoftweetingbeingcovid-19positiveisthedateofinfectionincasenootherinformationwereavailable.Figure2:Datacollectionsteps.

Afinallistof270covid-19userswereidentifiedamongst80weresharingtheirsymptomspublicly.Tofurtherunderstandthechronologicalorderofthesymptoms,

. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 12, 2020. .https://doi.org/10.1101/2020.06.10.20127225doi: medRxiv preprint

Page 5: Identifying and Ranking Common COVID-19 …...2020/06/10  · Twitter platform enables obtaining multiple features (such as age, sex, geo-location, … etc.) along with informative

weaskedthroughTwitterpersonalmessagesuserstoorderthefirstthreesymptomstheyexperiencedrightbeforeoraftertestedpositiveforcovid-19.Werecordthesymptomsfromfirsttolastbasedonthereceivedresponsesandwhatisavailablepubliclyontheusers’tweets.Incasenoorderwasgiven,animplicitorderisassumedfollowingtheorderofwhichthesymptomswerementionedbytheuser.Trackingtweetscontainingspecifickeywordsissimplynotenoughtohavethebigpictureaboutthediseasedynamics[9].Manypatientsdetailtheirexperiencewhileinfected,hence,knowingtheirhealthcondition,sentiment,andtrackingusefulinformationmayleadtoabetterunderstandingofthediseasesymptoms.Inparticular,wefoundtweetsthatwerepostedwithin±5daysofinfectiondatetocontainvaluableinformationaboutearlysymptoms,allowingustoprocessandrankthesymptoms.Figure3highlightsthreetweetsbythreedifferentcovid-19patientsthatindirectlyrelatesymptomsbeforeorafterdiagnosedwithcovid-19.Forsimplicity,wesetafakedate(28Apr,2020)forallthethreeusingtheTweetGenservice[10].User..1wastestedpositiveonApr26andtweetedonApr28aboutthelossofsmell.User…2testedpositiveMay1st,threedaysaftercomplainingaboutaheadache.User..3testedpositiveApr29th,onedayaftertweetingherwishtobeabletotastethefood.Figure3:Exampleoftweetscollectedwithin±5daysoftheusertweetingbeingcovid-19positive

TheexamplehighlightedinFigure3demonstratesthatminingtwitterforcovidsymptomsrequiremorethanasimplekeywordsearch.Inprinciple,thecontextofthetweet,narratedbyacovid-19patient,isalsoimportant.Therefore,itisimportanttolooknotonlytheverbatimofthetweetbutalsotoitscontext.Tobuildahigh-qualitydatabaseofcovidsymptomsbasedonArabictweets,wehavereliedonmanualsymptomsextraction.However,wehaveusedTwitterAPItoconstructa

. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 12, 2020. .https://doi.org/10.1101/2020.06.10.20127225doi: medRxiv preprint

Page 6: Identifying and Ranking Common COVID-19 …...2020/06/10  · Twitter platform enables obtaining multiple features (such as age, sex, geo-location, … etc.) along with informative

socialnetworkgraphforthe270usersandthesoftwareGephi[11]tovisualizetheresultedgraph.

Results ThemajorityofthecaseswererecordedinMay(78%,n=210),followedbyApril(14%,n=39)andMarch(8%,n=21).ThissurgeofMayreportsisunderstandableasmostofworldcountries,letalonetheArabicspeakingcountries,witnessedagreatincreaseinthenumberofconfirmedcases.Asforthedemography,usersfromSaudiArabia,Kuwait,andUAEconstitute85%ofreportswithSaudiArabiabeingthelargestclusterofreports(46%).Theothercountries(Egypt,Iraq,Bahrain,Qatar,UK,USA,Belgum,andGermany)togetherconstitutetheremaining15%.Needlesstosay,someoftheadoptedstrategiestopreventfurtherspreadofthevirus(e.g.,activescreeningbytheMinistryofHealthinSaudiArabia[12])mayalsohelpedinfindingmorereportsinMaycomparedtoothermonths.Wehavewitnessedthisfirsthandassomeoftheasymptomaticreportsweremainlyaresultofearlyactivescreening.ThefactthatalmosthalfofthereportscamefromSaudiArabiaisnotsurprisingasitsoneofthetopcountriesparticipatingonTwitterwithmorethan15millionusers[13].Wehavecollectedalmost900symptomsfromthe270reports(asshowninFigure4).ThedailynumberofcollectedtweetsisalsohighlightedinFigure5.

Figure4indicatesthatmostofthereportedcasesexperiencebetween2to5symptoms,whereas13%ofthereportedcaseswereasymptomatic.Table1liststhefrequencyofeachsymptomorderedfromthemostprevalentsymptomtotheleast.OnlyFeverexperiencedbymorethan50%ofthepatients.Thefrequencyofsymptomsappearstobeconsistentacrossmaleandfemalepatients(Corr.Coeff.=0.966).Further,Table2liststhetopthreemostreportedsymptoms.Feverandheadachewerecommonlythefirstreportedsymptoms.Thetop4symptomsthatcoincidewithfeverwereheadache(23.7%),cough(14.4%)anosmia(13.7%),andageusia(12.2%).Othersymptomshaverelativelylowerfrequencywithfever.Inaddition,Table3liststhetop-8commonsymptomsforSaudiArabiaandKuwaitwhichcorrespondforabout81.2%ofthereports.ThesymptomshaveaCorr.Coeff.of0.835betweenthetwocountries.Figure4:Numberofreportsdistributedbynumberofsymptoms.

. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 12, 2020. .https://doi.org/10.1101/2020.06.10.20127225doi: medRxiv preprint

Page 7: Identifying and Ranking Common COVID-19 …...2020/06/10  · Twitter platform enables obtaining multiple features (such as age, sex, geo-location, … etc.) along with informative

Figure5:NumberofdailycollectedreportsfromTwitterforMarchtoMay2020

Table1:Mostcommonsymptomsreportedbytheusers

Symptom Allusers n=234(%)

Malen=171(%)

Femalen=63(%)

Fever 139(59) 98(57) 41(65)Headache 101(43) 68(40) 33(52)Anosmia 91(39) 63(37) 28(44)Ageusia 72(31) 51(30) 21(33)Fatigue 68(29) 54(32) 14(22)

0

10

20

30

40

50

60

70

0 1 2 3 4 5 6 7 8 9 10

Num

berofReports

NumberofSymptoms

02468101214161820

14-Mar

21-Mar

28-Mar

04-Apr

11-Apr

18-Apr

25-Apr

02-May

09-May

16-May

23-May

Num

berofReports

Date

. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 12, 2020. .https://doi.org/10.1101/2020.06.10.20127225doi: medRxiv preprint

Page 8: Identifying and Ranking Common COVID-19 …...2020/06/10  · Twitter platform enables obtaining multiple features (such as age, sex, geo-location, … etc.) along with informative

Cough 62(26) 48(28) 14(22)SoreThroat 42(18) 30(18) 12(19)Dyspnea 33(14) 26(15) 7(11)Diarrhea 27(12) 22(13) 5(8)

RunnyNose 23(10) 17(10) 6(9)Arthralgia 16(7) 10(6) 6(9)ChestPain 15(6) 13(8) 2(3)BackPain 14(6) 11(6) 3(5)Anorexia 14(6) 11(6) 3(5)BodyAche 12(5) 8(5) 4(6)Nausea 12(5) 8(5) 4(6)

Osteodynia 11(5) 8(5) 3(5)Drythroat 9(4) 6(3) 3(5)Myalgia 9(4) 7(4) 2(3)Dizziness 8(3) 6(3) 2(3)Chills 7(3) 5(3) 2(3)

NasalCongestion 7(3) 4(2) 1(2)Sinusitis 7(3) 3(2) 4(6)

Table2:Top-8first,second,andthirdsymptomsreportedbytheusers

First Second ThirdFever Fever Fever

Headache Headache HeadacheAnosmia Fatigue AnosmiaFatigue Cough AgeusiaCough Ageusia Fatigue

Sorethroat Anosmia CoughRunnyNose SoreThroat AnorexiaDiarrhea Arthralgia Dyspnea

Table3:Top-8commonsymptomsforSaudiArabiaandKuwait

Symptom Saudi,n=110(%) Kuwait,n=80(%)Fever 65(59) 45(56)

Headache 42(38) 38(48)Anosmia 46(42) 21(26)Ageusia 36(37) 19(24)Fatigue 31(28) 19(24)Cough 21(19) 19(24)

SoreThroat 22(20) 11(14)Dyspnea 14(13) 11(14)

. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 12, 2020. .https://doi.org/10.1101/2020.06.10.20127225doi: medRxiv preprint

Page 9: Identifying and Ranking Common COVID-19 …...2020/06/10  · Twitter platform enables obtaining multiple features (such as age, sex, geo-location, … etc.) along with informative

Wehaveconstructedasocialgraphfortheuserstodiscoverinsightsrelatedcovid-19communitiesinsocialnetworksasshowninFigure6.AnedgebetweenXandYrepresentsthefactthatXfollowsYorYfollowsXorbothfolloweachotheronTwitter.SaudicaseswerecoloredingreenwhiletheKuwaitcasesareblueandorangeforotherplacesofresidence.Thegraphshows164nodeseachnodewithasizeproportionaltoitsdegreeinthegraph.Theremaininguserswerefoundtobedisconnectedfromallotherusersandthusremovedfromthenetwork.ItwasbuiltbythesoftwareGephi[11]andrelationswereextractedbyTwitterAPI.Clearly,Kuwaitusersaremoreconnectedcomparedtoothersandsomeformacliqueofthreeormoreusers.Figure6:Thesocialnetworkgraphoftheusers

. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 12, 2020. .https://doi.org/10.1101/2020.06.10.20127225doi: medRxiv preprint

Page 10: Identifying and Ranking Common COVID-19 …...2020/06/10  · Twitter platform enables obtaining multiple features (such as age, sex, geo-location, … etc.) along with informative

Discussion Thisworkidentifiedcommoncovid-19symptomsfromArabicpersonalreportsonTwitter.Suchstudycomplementsotherrecentstudies[5-6,9]thatwerefocusedonEnglishtweetsorspecificdemographicgroups.Thestudywascarriedinawaytoreportnotonlythesymptomsbuttheirtimelineasnarratedbyusers.Socialnetworkshavebecomethedefactocommunicationchannelforlargenumberofpeople.Manyindividualsaroundtheglobewrite,interact,orevenjustbrowsethecontentofthesocialnetworkscountlesstimesaday.Socialnetworkshavethepropertyofbeingcontinuouslyupdatedbynewinformationprovidedbyotherglobalcitizens.Assuch,itiscrucialtomonitortheircontenttoidentifyhealthissues[14-15].Onepotentialbenefitofanalyzingsocialnetworksisunderstandingcovid-19symptomsandidentifyingpeopleathighrisk[7].Anosmiabeingoneofthetopthreereportedsymptoms,mentionedin39%ofthereports,wasasurprisingresultinourstudy.Severaltweetscomplainedhowlongitlastedbeforetheystartedtosmellagain.Oursamplesizeisstillrelativelysmalltomakeanygoodjudgementinthisregard.However,recentclinicalstudieshavereportedfindingAnosmiain35.7%ofthemildcasesofcovid-19[16]whichisrelativelyclosetoourestimationfromArabictweets.Indeed,thesizeofself-reportsreflectthetestingcapacityindifferentcountries.AsofJune9,2020,SaudiArabiahasdonealmostonemilliontestsandKuwaithasroughlyexceeded350,000tests[17].

Limitations ThereportedcasesfromEgypt,theArabiccountrywithlargestpopulation,wereinadequatelyrepresentative.Thiscouldbeattributedtofactorssuchasthepreferredsocialplatform(e.g.,Facebook),thedialectanduseoflocalidioms.OurstudytrackedtwowidelyusedkeywordstoidentifyArabiccovid-19patientsonTwitterthenmanuallyextractsymptoms.Morecomplexkeywordscouldrevealmoreinterestingpatternsaboutsymptoms,andthisbringstheneedtoestablishacomprehensivemedicaldictionaryfordifferentlocalArabicdialect.ThedictionarycanbeutilizedwhenminingdifferenthealthopinionsandconditionsfromtheArabiccontentinsocialnetworks.Ourmainmotivationistoextractsymptomsfromuserswhoarelikelytookthediseasetestand,hence,tweetedbasedonitsresult.Inthisstudy,wehavenotusedothercovid-19sources.Specifically,studyingtheArabiccontentofpersonalreportsfrombothFacebookandTwitterwillenrichthestudy.

. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 12, 2020. .https://doi.org/10.1101/2020.06.10.20127225doi: medRxiv preprint

Page 11: Identifying and Ranking Common COVID-19 …...2020/06/10  · Twitter platform enables obtaining multiple features (such as age, sex, geo-location, … etc.) along with informative

ThenoticeableincreaseinMayreportscomparedtoothermonthsshowtheimportanceofdevelopingareal-timesurveillancesystembasedonthesymptomsreportedbytheArabiccontentofTwitter.Italsosuggestsfurtherstudiesintotheinformationsharingbehavior[19]indifferentcommunitiesand.

Conclusion Thecollectedtweetsofsymptomsprovidedsomeinsightsintocovid-19symptomsandtheirchronologicalorder.WehaveshownthemostcommonsymptomsassociatedwithArabicself-reportsofsymptoms.Wehaveanalyzedthefirst,second,andthirdcommonsymptomexperiencedbytheusers.Furthermore,weanalyzedsymptomsprevalenceinthetwolargestclustersfoundinourdatabase:SaudiArabiaandKuwait.

Acknowledgements

ThisworkwassupportedbyKingAbdulazizCityforScienceandTechnology(GrantNumber:5-20-01-007-0033).

Authors' Contributions EisaAlanaziandAbdulazizAlashaikhdesignedthestudyandwrotethemanuscript.SarahAlqurashidevelopedthesocialnetworkanalysispartandcollectedrelatedtweetsfromtheTwitterAPI.AuedAlanaziextractedandtranslatedthesymptomsfromthecollectedpersonalreportstotheirscientificnames.Allauthorsapprovedthefinalversionofthemanuscript.

Conflicts of Interest Nonedeclared

1. Kathy Lee, Ankit Agrawal, Alok Choudhary, “Real-time disease surveillance using Twitter data: demonstration on flu and cancer,” KDD '13: Proceedings of the 19th ACM SIGKDD international conference on Knowledge discovery and data mining, August 2013, Pages 1474-1477

2. B. Alkouz and Z. A. Aghbari, "Analysis and prediction of Influenza in the UAE based on Arabic tweets," 2018 IEEE 3rd International Conference on Big Data Analysis (ICBDA), Shanghai, 2018, pp. 61-66, doi: 10.1109/ICBDA.2018.8367652.

3. B. Alkouz, Z. A. Aghbari and J. H. Abawajy, "Tweetluenza: Predicting flu trends from twitter data," in Big Data Mining and Analytics, vol. 2, no. 4, pp. 273-287, Dec. 2019, doi: 10.26599/BDMA.2019.9020012.

4. Alotaibi, Shoayee and Mehmood, Rashid and Katib, Iyad and Rana, Omer and Albeshri, Aiiad , “Sehaa: A Big Data Analytics Tool for Healthcare Symptoms

. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 12, 2020. .https://doi.org/10.1101/2020.06.10.20127225doi: medRxiv preprint

Page 12: Identifying and Ranking Common COVID-19 …...2020/06/10  · Twitter platform enables obtaining multiple features (such as age, sex, geo-location, … etc.) along with informative

and Diseases Detection Using Twitter, Apache Spark, and Machine Learning,” Applied Sciences, vol. 110, no. 4, 2020, https://doi.org/10.3390/app10041398

5. Abeed Sarker, Sahithi Lakamana, Whitney Hogg Bremer, Angel Xie, Mohammed AliAl-Garadi, Yuan-Chi Yang: Self-reported COVID-19 symptoms on Twitter: An analysis and a research resource. medRxiv 2020.04.16.20067421; doi:https://doi.org/10.1101/2020.04.16.20067421

6. Ari Z. Klein, Arjun Magge, Karen O’Connor, Haitao Cai, Davy Weissenbacher, Graciela Gonzalez-Hernandez, “A Chronological and Geographical Analysis of Personal Reports of COVID-19 on Twitter,” medRxiv 2020.04.19.20069948; doi: https://doi.org/10.1101/2020.04.19.20069948

7. Sinnenberg L, Buttenheim AM, Padrez K, Mancheno C, Ungar L, Merchant RM. Twitter as a Tool for Health Research: A Systematic Review. Am J Public Health. 2017 Jan;107(1):e1-e8. doi: 10.2105/AJPH.2016.303512. Epub 2016 Nov 17. PMID: 27854532; PMCID: PMC5308155.

8. Mawid Service https://www.moh.gov.sa/en/eServices/Pages/cassystem.aspx [accessed 27-05-2020]

9. Mackey T, Purushothaman V, Li J, Shah N, Nali M, Bardier C, Liang B, Cai M, Cuomo R Machine Learning to Detect Self-Reporting of Symptoms, Testing Access, and Recovery Associated With COVID-19 on Twitter: Retrospective Big Data Infoveillance Study JMIR Public Health Surveill 2020;6(2):e19509 URL: http://publichealth.jmir.org/2020/2/e19509/ doi: 10.2196/19509

10. TweetGen https://www.tweetgen.com/ [accessed: 02-06-2020] 11. Gephi The Open Graph Viz Platform https://gephi.org [accessed: 05-06-2020] 12. Saudi Arabia’s active mass testing contains COVID-19 spread

https://www.arabnews.com/node/1662856/saudi-arabia [accessed 25-05-2020] 13. Statista: Leading countries based on number of Twitter users as of April 2020.

https://www.statista.com/statistics/242606/number-of-active-twitter-users-in-selected-countries/ [accessed 28-05-2020]

14. Paul MJ, Sarker A, Brownstein JS, Nikfarjam A, Scotch M, Smith KL, Gonzalez G. Social media mining for public health monitoring and surveillance. InBiocomputing 2016: Proceedings of the Pacific symposium 2016 (pp. 468-479).

15. Chunara, R., Bouton, L., Ayers, J.W. and Brownstein, J.S., 2013. Assessing the online social environment for surveillance of obesity prevalence. PloS one, 8(4).

16. Ruth Levinson, Meital Elbaz, Ronen BenAmi, David Shasha, Tal Levinson, Guy Choshen, Ksenia Petrov, Avi Gadoth, Yael Paran “Anosmia and dysgeusia in patients with mild SARS-CoV-2 infection” medRxiv 2020.04.11.20055483v1doi: https://doi.org/10.1101/2020.04.11.20055483.

17. COVID-19 Pandemic https://www.worldometers.info/coronavirus/ [accessed 09-06-2020]

18. Neiger BL, Thackeray R, Burton SH, Thackeray CR, Reese JH 19. Use of Twitter Among Local Health Departments: An Analysis of Information

Sharing, Engagement, and Action J Med Internet Res 2013;15(8):e177 URL: https://www.jmir.org/2013/8/e177 DOI: 10.2196/jmir.2775 PMID: 23958635 PMCID: PMC3758023

. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 12, 2020. .https://doi.org/10.1101/2020.06.10.20127225doi: medRxiv preprint

Page 13: Identifying and Ranking Common COVID-19 …...2020/06/10  · Twitter platform enables obtaining multiple features (such as age, sex, geo-location, … etc.) along with informative

Abbreviations API:applicationprogramminginterfaceCOVID-19:coronavirusdisease

. CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 12, 2020. .https://doi.org/10.1101/2020.06.10.20127225doi: medRxiv preprint