• No results found

Prevalence, survival analysis and multimorbidity of chronic diseases in the general veterinarian-attended horse population of the UK

N/A
N/A
Protected

Academic year: 2021

Share "Prevalence, survival analysis and multimorbidity of chronic diseases in the general veterinarian-attended horse population of the UK"

Copied!
9
0
0

Full text

(1)

ContentslistsavailableatScienceDirect

Preventive

Veterinary

Medicine

jou rn al h om ep a g e :w w w . e l s e v i e r . c o m / l o c a t e / p r e v e t m e d

Prevalence,

survival

analysis

and

multimorbidity

of

chronic

diseases

in

the

general

veterinarian-attended

horse

population

of

the

UK

Claire

E.

Welsh

a,∗

,

Marco

Duz

b

,

Timothy

D.H.

Parkin

a

,

John

F.

Marshall

a

aEquineClinicalSciencesDivision,WeipersCentreEquineHospital,SchoolofVeterinaryMedicine,UniversityofGlasgow,UK bSchoolofVeterinaryMedicineandScience,UniversityofNottingham,UK

a

r

t

i

c

l

e

i

n

f

o

Articlehistory: Received6January2016

Receivedinrevisedform14June2016 Accepted24July2016 Keywords: Horse Chronic Veterinary Survival

a

b

s

t

r

a

c

t

Theaverageageoftheglobalhumanpopulationisincreasing,leadingtoincreasedinterestintheeffects ofchronicdiseaseandmultimorbidityonhealthresourcesandpatientwelfare.Ithasbeenpositedthat theaverageageofthegeneralveterinarian-attendedhorsepopulationoftheUKisalsoincreasing,and thereforeitcouldbeassumedthatchronicdiseasesandmultimorbiditywouldposeanincreasingriskhere also.However,evidenceforthistrendinageingisverylimited,andthecurrentprevalenceofmanychronic diseases,andofmultimorbidity,isunknown.Usingtextminingoffirst-opinionelectronicmedicalrecords fromsevenveterinarypracticesaroundtheUK,Kaplan-MeierandCoxproportionalhazardmodelling, wewereabletoestimatetheapparentprevalenceamongveterinarian-attendedhorsesofninechronic diseases,andtoassesstheirrelativeeffectsonmedianlifeexpectancyfollowingdiagnosis.Withthese methodswefoundevidenceofincreasingpopulationage.Multimorbidityaffected1.2%ofthestudy population,andhadasignificanteffectuponsurvivaltimes,withco-occurrenceoftwodiseases,and threeormorediseases,leadingto6.6and21.3timesthehazardratiocomparedtonochronicdisease, respectively.Laminitiswasinvolvedin74%ofcasesofmultimorbidity.Thepopulationofhorsesattended byUKveterinariansappearstobeaging,andchronicdiseasesandtheirco-occurrencearecommon features,andassuchwarrantfurtherinvestigation.

©2016TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

Thenon-racing,mixedbreedhorsepopulationoftheUKhas

beenrelativelyneglectedtodateintermsofresearchintodisease

burdenand populationstructure (Melloret al.,1999;Hotchkiss

etal.,2007;Irelandetal.,2013;Wylieetal.,2013).Theremaybeas

manyasonemillionhorsesintheUK,yetlittleisknownaboutthe

healthandlongevityofthevastmajority(BritishEquestrianTrade

Association,2011).Reasonsforthisincludethelackofacentral

databaseofmandatoryhorsepassportdetailssincewithdrawalof

fundingfromtheNationalEquineDatabasein2012,andthe

frag-mentednatureoffirst-opinionequineveterinaryservices(Boden

etal.,2013).Collationofhealthanddiseasedataoccursonasmall

scaleperiodically,butreliesuponeitherowner-reportingofhealth

problemsorvoluntarysubmissionoflaboratorysamples,which,

althoughinformativeandhelpful, arebothsubjecttosignificant

biasifresultsweretobeextrapolatedtothegeneralequine

popu-lation(Slater,2014;AnimalHealthTrust(AHT)etal.,2015).Chronic

∗ Correspondingauthor.

E-mailaddress:Claire.Welsh@Glasgow.ac.uk(C.E.Welsh).

orrecurrentdiseaseisundoubtedlyafeatureofthispopulation,but

theextenttowhichitaffectslongevityandwelfareisunknown.

Multimorbidity is defined as the co-occurrence of multiple

diseaseconditionswithinanindividualwithoutreferencetoan

indexcondition(Islametal.,2014).Thestudyofmultimorbidity

in medical research is currently popular, as recognitionof the

greaterstrainmultimorbidityplacesuponhealthservices,andits

detrimentaleffectonpatientwelfareand longevityaregrowing

(Islametal.,2014;Pacheetal.,2015).Understanding

multimor-biditypatternscan helpuncover commonalitiesin aetiologyor

risk factors betweenpreviously unlinked conditions, and could

improvepatientcarebyshiftingtreatmentemphasisfrom

indi-vidualdiseasestoco-occurringgroupsofdisease,especiallywhere

treatmentregimensforco-occurringdiseasesarediscordant.The

averagehumanlifespanintheUKcontinuestorise,andwithit,

the proportion of people living withmultiple chronic diseases

(Pacheetal.,2015;Wohlandetal.,2015).Ithasbeenoftencited

thattheaverageageofhorsesintheUKisalsoincreasing,butthis

assumptionisbaseduponlimitedstudies,andmethodologiesless

robustthanthoseusedbyUKhumanhealthauthorities(Brosnahan

andParadis,2003).

Thisstudyaimedtousefirst-opinionmedicalrecordsto

quan-tifytheburdenofanumberofcommonchronicdiseaseswithin

http://dx.doi.org/10.1016/j.prevetmed.2016.07.011

0167-5877/©2016TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4. 0/).

(2)

asampleofthegeneralveterinary-attendedhorsepopulationin

theUK,and toestimatetheireffects upon survival. The

appar-entprevalenceofcomorbidity(co-occurrenceoftwodiseases)and

multimorbidity(co-occurrenceofmorethantwodiseases)were

alsosought,asweretheeffectsofco-andmultimorbidstateson

survival.Asecondaryaimwastoimplementdemographic

method-ologiestoascertainwhetherthepopulationunderstudywasageing

overtime,ashasbeenpreviouslypostulated.Theseresultswill

con-stitutethefirstsuchanalysesonthislargeanddiversepopulation,

andwillserveasbenchmarksforfuturestudiesoftheeffectsof

chronicdiseaseandmultimorbidityinhorses.

2. Materialsandmethods

2.1. Data

Free-text first-opinion equine medical records spanning

twenty-six years (1987–2013) were collected from a

conve-niencesampleofsevenfirstopinionequineveterinarypractices

around the UK, and reformatted into comma-separated values

format (n=1001375 records). Records included the following

database;uniquenumericalidentificationnumber,dateofentry

into the system (date of record), sex, date of birth, breed,

practice, and a field containing free-text medical records with

prescription information. Records missing the date of birth or

record entry were excluded, as were records of non-equine

consultations, and records entered after 2012. Sex was

con-verted into four categories; unknown, female, male, neutered

male.Breedwasconvertedintothefollowinggroups;Arab/Arab

cross,Cob/Cobcross,Draught/Draughtcross,Native/Nativecross,

pony/pony cross, Thoroughbred/Thoroughbred cross,

Warm-blood/Warmbloodcross,Welsh/Welshcross,otherandunknown.

DescriptionsofeachveterinarypracticearecontainedinTable1.All

datacleaningwasconductedinRstatisticalenvironment(RCore

DevelopmentTeam,2015).

2.2. Textmining

Ninechronic(orrecurrent)conditionswerechosenforstudya

priori,afterdiscussionwithexperiencedequineclinicians;

neopla-sia,pituitaryparsintermediadysfunction(PPID),equinemetabolic

syndrome(EMS), grass sickness, laminitis, navicular syndrome,

osteoarthritis,recurrentairwayobstruction(RAO),andsarcoids.

Thesewerechosenasdiseasesthatcanbediagnosedinfirstopinion

practice,whicharechronicorrecurrentinnatureandare

com-monlyincurable.Theywerealsoconsideredtohavenamesthat

wouldbeunlikelytoappearinmedicalrecordsunlesstheanimal

inquestionwasthoughttobesufferingfromthem.Commercially

availabletextminingsoftware(WordStatv7.0,ProvalisResearch

Inc.)wasusedtoconstructdictionariesthatwouldallow

identifica-tionofrecordswhereeachdiseasewasmentioned,inaniterative

processmodifiedfromtheapproachdetailedbyLametal.(Lam

etal.,2007)andvalidatedbyAnholtetal.(2014).‘Diagnosis’of

dis-easeinthiscontextwasdefinedasinclusionofthesekeywords

orphrases,withoutrequirementofadditionalinformationsuchas

testresults. ‘Diagnoses’werethereforereflective ofthethought

processandprofessionalopinionoftheveterinarianregardingthe

caseathand,asopposedtoanaccurateandverbosedescriptionof

thediagnosticcriteriaofthecase(whichwerecommonlymissing).

Additionally,adictionarywasconstructedtoidentifyrecordsof

death(naturaloreuthanasia)(seeAppendixA).Thesecriteriawere

choseninordertoidentifythereal-worlddecisions(e.g.

euthana-sia)takenfollowingdeliveryofaveterinarian’sclinicaljudgement

ineachcase. Dictionaryconstructionwasperformedonthefull

datasettomaximisecoverage,andnegationswereaddedtoalist

ofexcludedwordsandphrasesasnecessary.Internalvalidationwas

conductedthrough manuallycheckingthe‘keyword-in-context’

outputforeachdisease;andalteringthedictionaryand/orexcluded

wordslistasnecessary. Data weresubsequently exportedwith

onecolumnperdisease(containingbinarydataindicating

pres-ence/absenceofdisease)inplaceoffreetextrecords;plusabinary

columnindicatingpresence/absenceofdeathateachrecord.

2.3. Lifeexpectancyfollowinginitialveterinarycareepisode

Periodlifetableswereconstructedforthree-yearperiodsnear

thebeginningandendofthestudy(1995–1997,and2010–2012),

forrecordsfromfivepractices(twopracticeswereexcludeddueto

significantlyloweragesatthetimeofdeath[Practice4],andhigh

annualproportionofdeaths[Practice7]).Lifeexpectancy

(follow-ingfirstrecordedveterinaryepisode)with95%confidenceintervals

wascalculated(Chiang,1968,1978).

2.4. Univariableandmultivariablemodellingofchronicdiseases

SurvivalanalyseswereconductedusingRpackages‘survMisc’

and‘survival’(Dardis,2015;Therneau,2015).Foreachhorsethat

hadatleastonerecordofchronicdisease,thetimefromdiagnosis

toitslastrecordinthedatawascomputed,andthepresenceof

arecordofdeathwasrecordedas0/1.Similarly,forthosehorses

thatwerenotdiagnosedwithanyoftheselecteddiseases,thetime

fromtheirfirstentryinthedata,totheirlast,wascomputed,and

whethertheywererecordedashavingdiedwasnoted.

Kaplan-Meiersurvivalanalysiswasperformedtoestimatemediansurvival

timefollowingdiagnosis,orfollowingentryintothedataset(inthe

caseofnochronicdisease).UnivariableCoxproportionalhazard

modelswereconstructedtoassessthesignificance oftheeffect

of disease,sex, breed,practice,and ageat thetime of

diagno-sis (years)upon survival. Variables wereretained for potential

inclusioninthemultivariablemodelifthep-value<0.25,andthe

likelihoodratiotest(LRT)p-valuecomparedwiththenullmodel

was<0.05.Categoricalvariableswerere-groupedbasedon

similar-ityofexponentiatedcoefficientsandWaldsignificance,asrequired.

Aforwardstepwisemanualmodelbuildingprocedurewas

imple-mented,withsignificantvariablesincludedinorderofhighestto

lowestlog-likelihood(Dohoo,2009).Variableswereretainedinthe

multivariablemodelifLRTp-value<0.05.Allpairwiseinteractions

betweenretainedvariableswereinvestigatedforsignificanceand

theireffectonexistingcoefficients.Proportionalhazards

assump-Table1

Descriptivedetailsofaconveniencesampleofsevenfirst-opinionveterinarypracticesservinghorses,thatcontributeddatatothecurrentstudy.

Practice Numberofveterinarians Location Provideownout-of-hours RCVSAccredited Speciescovered Numberofbranches

1 11 Scotland Yes Yes Mixed 2

2 21 CentralEngland Yes Yes Equine 1

3 17 NorthernEngland Yes Yes Mixed 5

4 14 CentralEngland Yes No Mixed 4

5 11 SouthernEngland Yes No Equine 1

6 4 NorthernEngland Yes Yes Large 1

7 8 NorthernEngland Yes No Mixed 2

(3)

Fig.1.Distributionsofage(years)atthetimeofdeathofhorsesseenataconveniencesampleofUKfirst-opinionveterinarypracticesbetween1987and2013.

tionsofthemodelglobally,andofeachvariableindividually(all levelsofcategoricalvariablesassessedenmasseandindividually) wereassessedbyvisualinspectionofscaledSchoenfeldresiduals (Therneau andGrambsch,2000;Therneau, 2015).Theinfluence

of individualobservations wasassessed byvisual inspectionof

delta-betas forcontinuous predictors (Therneauand Grambsch,

2000).

2.5. Modellingmultimorbidity

For the purposesof this study,‘comorbid’ indicatesa horse

diagnosedwithtwochronicdiseases,and‘multimorbid’indicates

a horse diagnosed with more than two diseases. The

associa-tion of disease co-occurrence with survival was analysed after

computingthetimefromcommencementofthelifetimedisease

co-occurrencestate(i.e.‘comorbid’horses:dateofdiagnosisofthe

seconddisease,‘multimorbid’horses:dateofdiagnosisofthethird

disease)tothedateofthelastentryforthathorse,with‘nochronic

disease’asthereferentlevel.Modelbuildingproceededsimilarlyto

diseasemodellingabove.Independentvariablesscreenedfortheir

associationwithsurvivalincludedlifetimediseaseco-occurrence

status(none,one,two,ormore thantwochronicdiseases),age

attimeofdiagnosis(years),practice,breedandsex.Again,breeds

weregroupedaccordingtosimilarityofexponentiatedhazardratio

and Waldsignificance. Kaplan-Meier survivalanalysis was

con-ductedandprobabilityofsurvivalplotted.

Ethicalapproval for this studywasgranted by theResearch

Ethics Committee,School of VeterinaryMedicine, University of

Glasgow.

3. Results

3.1. Demographiccharacteristics

The cleaned dataset contained 515807 records from 70477

horses, of which 12796 (18%) were male, 25214 (36%) were

neuteredmales,26018(37%) werefemale,and 6487(9%)were

Table2

Summaryoftheage(years)atthetimeofdiagnosisforaselectionofchronicdiseases recordedinaUKdatabaseofelectronicmedicalrecordsfromthegeneralhorse populationbetween1987and2013.

Disease Minimum Q1 Median Q3 Maximum

Nochronicdisease 0 3 6 12 40 Osteoarthritis 0 7 14 21 39 Laminitis 0 8 15 22 40 PPID 0 15 21 26 40 EMS 1 15 21 28 38 GrassSickness 0 5 9 15.5 34 NavicularSyndrome 0 6 10 16.5 36 Neoplasia 0 7 14 22 40 RAO 0 6 13 21 39 Sarcoids 0 6 11 19 40 Q125thpercentile;Q375thpercentile.

ofunknownsex.Sevenpracticescontributedtothedataset.The annualageatthetimeofdeathforhorsesineachpracticeisshown

inFig.1,andtheproportionofhorsesthatdiedannuallyisshown

inFig.2.Overall,themeanageatdeathforallhorseswas16.5years

(median17,IQR8–24years).

3.2. Quantificationofdiseaseburden

Intotal,9013horses(12.8%)hadatleastonechronicdisease

(maximumof6)overthecourseofthestudy.Thenumberofrecords

inwhicheachdiseasewasnoted,andthenumberofhorsesthose

recordspertainto(inbrackets),wasasfollows;neoplasia 1159

records(978horses),PPID3199(2011),EMS154(140),grass

sick-ness72(66),laminitis6110(4082),navicularsyndrome67(60),

osteoarthritis1281(1171),RAO1388(1232),sarcoids3476(2070).

Deathwasrecordedforatotalof4689(6.7%)horsesoverthestudy

period.

Asummaryoftheageatthetimeofdiagnosisforeachdisease

(4)

Fig.2.Proportionofveterinary-attendedhorsesseenataconveniencesampleoffirst-opinionUKequineveterinarypracticesthatdiedorwereeuthanasedperpracticeper yearbetween1987and2013.

Fig.3.Lifeexpectancy(years)followingfirstrecordedveterinarycareepisodeofhorsesbornin1995–1997,versusthosebornin2010–2012,fordifferentagegroups.Grey bandsindicate95%confidenceintervals.Practice4andPractice7wereexcludedfromtheseanalysesduetosignificantdifferencesindemography.

3.3. Lifeexpectancyfollowingfirstveterinarycareepisode

Lifeexpectancyfollowingfirstveterinarycareepisodeforhorses

bornin 1995–1997and 2010–2012,separated byagegroup, is

showninFig.3.Horsesborninthelatertimeperiodwerefound

tohavea longerlifeexpectancyatbirththanthoseborninthe

earliertimeperiod.

3.4. Multimorbidityprevalence

Intotal,2125horseswerediagnosedwithmorethanonechronic

disease(3.0%ofthetotalpopulation;23.6%ofthosewithatleast

onechroniccondition).Ofthese,1582hadtwodiseases,406had

threediseases,120hadfourdiseases,16hadfivediseases,and1

hadsixdiseases.Informationonthenumberofhorsesexperiencing

eachcombinationofatleasttwoco-occurringdiseasesiscontained

inTable3.Laminitiswasafeatureof74%ofhorseswithmultiple

diseases,andPPIDwasafeaturein49%.

3.5. Survivalanalyses

Thefinalmultivariablemodelcontainedthefollowing

signifi-cantvariables;disease,ageatdiagnosis(years),breed,andpractice

(includedasarandomeffect).Kaplan-Meierplotsofsurvival

fol-lowingdiagnosisofeachdisease,andforhorseswithoutchronic

disease,areshowninFig.4.Mediansurvivaltimesestimatedfrom

thismodel,alongsideresultsofCoxproportionalhazardmodelling

(5)

Table3

Co-occurrencematrixofaselectionofchronicdiseasesinaUKgeneralveterinary-attendedhorsepopulation,identifiedthroughtextminingelectronicmedicalrecordsfrom sevenfirst-opinionpracticesbetween1987and2013.Numbersonthediagonalarethenumber(%)ofhorsesdiagnosedwiththatdiseaseinthedata(n=70477).

Osteoarthritis Navicularsyndrome Neoplasia PPID EMS GrassSickness Laminitis RAO Sarcoids Osteoarthritis 1171(1.7) Navicularsyndrome 3 60(0.1) Neoplasia 65 0 978(1.4) PPID 132 3 124 2011(2.9) EMS 14 0 13 86 140(0.2) GrassSickness 1 0 3 8 1 66(0.1) Laminitis 304 11 235 806 103 9 4082(5.8) RAO 79 6 68 136 10 4 321 1232(1.7) Sarcoids 121 4 166 256 23 3 408 148 2070(2.9)

EMSEquineMetabolicSyndrome;PPIDPituitaryParsIntermediaDysfunction;RAORecurrentAirwayObstruction.

Fig.4. Kaplan-MeiersurvivalplotofaselectionofchronicdiseasesfoundtosignificantlyaffecttheprobabilityofsurvivalinaUKveterinary-attendedgeneralhorsecohort.

Table4

ResultsofmultivariableCoxproportionalhazardmodelsofninechronicdiseasesdiagnosedinaconveniencesampleoffirst-opinionequineveterinarypracticesbetween 1987and2013,asdetectedthroughtextminingofelectronicmedicalrecords.

Numberofhorses HR(SE) Waldp-value HR95%CI

Disease Nochronicdisease 61464 1(REF)

Neoplasia 978 6.52(0.07) <0.01 5.72–7.44 PPID 2011 9.27(0.05) <0.01 8.37–10.27 EMS 140 9.61(0.17) <0.01 6.85–13.47 GrassSickness 66 18.74(0.18) <0.01 13.25–26.50 Laminitis 4082 5.94(0.04) <0.01 5.49–6.42 Navicularsyndrome 60 5.19(0.29) <0.01 5.72–7.44 Osteoarthritis 1171 5.08(0.07) <0.01 4.47–5.77 RAO 1232 4.51(0.07) <0.01 3.95–5.15 Sarcoids 2070 4.91(0.06) <0.01 4.39–5.49

Ageatdiagnosis(years) 1.10(0.02) <0.01 1.09–1.10

BreedGroup Other/unknown 67680 1(REF)

Arab 988 0.88(0.12) 0.28 0.70–1.11

Pony/Welsh 4632 0.79(0.06) <0.01 0.70–0.79

HR:HazardRatio,95%CI:95%confidenceinterval,SE:standarderror,REF:referentcategory.

ofhazards,asassessedusingplotsofscaledSchoenfeldresiduals, wasupheldforeachvariable.Noobservationswerefoundtoexert undueinfluenceonthemodel.

TheresultsofCoxproportionalhazardmodellingoflifetime dis-easeco-occurrencestatus(withnochronicdiseaseasthereferent level,n=61464)arecontainedinTable6.

Survival curves of lifetime disease co-occurrence status are

showninFig.5.

Theproportionofhorsesdiagnosedwithtwoormorechronic

diseases(multimorbidstatus)overtimeis showninFig.6.This

proportionwas0.5%in1995–1997,and1.3%in2010–2012.

4. Discussion

This study is the first to report estimates of median

(6)

Fig.5.Kaplan-Meiersurvivalplotoflifetimediseaseco-occurrencestatus(consideringaselectionofninechronicdiseases)ofahorsesseenataconveniencesampleofUK first-opinionequineveterinarypracticesbetween1987and2013.

Table5

Medianlifeexpectancyfollowingdiagnosis(otherindependentvariablesheldat theirreferenceormedianlevels)ofaselectionofchronicdiseasesinaconvenience sampleofthegeneralveterinary-attendedhorsepopulationoftheUKbetween1987 and2013.

Diagnosis Medianlifeexpectancyinyears followingdiagnosis(95%CI)

Median(IQR)ageat diagnosis(years) Nochronicdisease 22.7(21.9–NA) 6(3–12) Neoplasia 12.3(11.4–13.6) 14(7–22) PPID 9.8(9.0–10.7) 21(15–26) EMS 9.0(7.0–12.4) 21(15–28) GrassSickness 4.0(2.3–6.8) 9(5–15.5) Laminitis 13.4(12.7–14.3) 15(8–22) Navicularsyndrome 14.4(10.5–19.5) 10(6–16.5) Osteoarthritis 14.5(13.3–15.9) 14(7–21) RAO 16.4(15.0–17.9) 13(6–21) Sarcoids 15.4(14.3–16.7) 11(6–19) CIconfidenceinterval;NAestimatenotavailable.

IQRInter-quartileRange;EMSEquineMetabolicSyndrome;PPIDPituitaryPars IntermediaDysfunction;RAORecurrentAirwayObstruction.

Fig.6. AnnualproportionofhorsesseenataconveniencesampleofUKfirst-opinion veterinarypracticesthatwerediagnosedwithmorethantwooftheninespecified chronicdiseasesoverthestudyperiod(1987–2013).

Table6

ResultsofmultivariableCoxproportionalhazardmodelsofco-andmultimorbidityofninechronicdiseasesinadatabaseofelectronicmedicalrecordsfromaconvenience sampleofsevenUKequineveterinarypracticesbetween1987and2013.

Numberofhorses HR(SE) p-value HR95%CI Lifetimediseaseco-occurrencestatus Nochronicdisease 61464 1(REF)

Onechronicdisease 8201 3.06(0.06) <0.01 2.70–3.47 Twochronicdiseases 743 6.61(0.17) <0.01 4.70–9.31 Morethantwochronicdiseases 78 21.29(0.49) <0.01 8.13–55.77

Ageatdiagnosis(years) 1.08(0.01) <0.01 1.06–1.09

Practice 1 2889 1(REF) 2 38704 0.54(0.15) <0.01 0.40–0.73 3 1442 2.57(0.24) <0.01 1.60–4.11 4 14339 2.22(0.17) <0.01 1.58–3.11 5 5565 0.57(0.20) <0.01 0.38–0.85 6 5052 0.83(0.20) 0.35 0.57–1.22 7 2485 1.93(0.20) <0.01 1.32–2.83

Ageatdiagnosis*Lifetimediseaseco-occurrencestatus <0.01*

Ageatdiagnosis*Practice <0.01*

HR:HazardRatio,95%CI:95%confidenceinterval,SE:standarderror,REF:referentcategory.

(7)

chronicdiseasesintheveterinarian-attendedgeneralhorse

popu-lationoftheUK.Over12%ofthehorsesinthisstudywerediagnosed

withatleastoneofninechronicdiseases,and1.2%ofhorseswere

diagnosedwithmorethanonechronicdisease.

NotallhorsesintheUKareregisteredwithaveterinarian,and

ofthosethatare,aproportiondonotreceiveveterinaryproducts

orattention,thereforeusingelectronicmedicalrecordstoascertain

theprevalenceofdiseaseordemographicinformationisliabletoa

degreeofselectionbias.Theresultsreportedherecannot,therefore,

bedirectlyextrapolatedtothewider veterinary-attended

popu-lationofhorses.Similarly,duetotheconveniencenatureofthe

samplingofveterinarypracticescarriedout,theresultspresented

maynotbefullygeneralizabletothegeneralveterinary-attended

horsepopulationoftheUK.Currently inveterinarymedicinein

theUK,nosystematicfixedvocabularyexistsfortherecordingof

medicalevents,soclinicalsigns,differentialdiagnoses,testsand

treatmentsareallrecordedusingfree-textentries.Variabilityin

spelling,nomenclatureandphraseologyleadtoaplethoraofways

inwhichdiseaseisdocumented.Theiterativenatureofthe

text-miningprocess,where includedwordsorphrasesareinternally

validatedthroughinspectionoftheirappropriatenessincontext,

hasbeenshowntoleadtoacceptablelevelsofsensitivityand

speci-ficitycomparedwithmanualclassification(Anholtetal.,2014).

Categorisationerrorsareunavoidablewithanyformofautomated

text-mining,andthismustbeborneinmindwithregardstothe

resultspresentedhere,especiallywherenomenclature

necessar-ilyoverlapsbetweenrelateddiseases(e.g.sarcoids,PPIDandother

neoplasias).Additionally,ownersarefreetoregistertheiranimal

withanynumberofveterinarypractices,andtomovepracticesat

anytime.Therefore,therecordsusedinthisstudymaynot

repre-senttheentiretyofthatanimal’smedicalhistory,andduplication

(whereahorsereceivedmedicalattentionatmorethanoneofthe

practicesusedinthestudy)mayhaveoccurred,althoughthiscould

beassumedtoberare.Inaddition,dictionarytermswerechosen

tooptimisespecificityofcaseidentification,withsome

unavoid-ablereductioninsensitivity,inordertobesureofthediagnoses

assigned,thusdisease‘prevalence’reportedheremaybelessthan

reportedinotherstudies,forsomediseases.Therewaspotentialfor

adegreeofoverlapinthedictionariesusedtoidentifycasesof

sar-coidsandnon-sarcoidneoplasia,howeverthetermswereretained

duetothestrongsuspicionamongsttheauthorsthatthesewould

beaccuratelydifferentiatedinmostcasesbyveterinarians,dueto

theprevalenceandcharacteristicnatureofsarcoids.

Thegroupofdiseasesstudiedintheseanalysesrepresentasmall

numberofchronic/incurableconditionsinequinemedicine,but

theybearcertainsimilaritiesintermsofnomenclature.Inclusion

ofthediseasename(oraderivationofit)inananimal’smedical

recordswasconsideredtobeverylikelytoindicatethatthe

vet-erinarianenteringtherecordwasstronglysuspicious of,orhad

diagnosed,thatcondition.Thisisunlikeconditionssuchasdental

disease,chroniclamenessorrecurrentcolic,whoseterminologyis

variable,andcouldrelatetoaspectrumofdiseaseincludingcurable

conditions.Thissuppositionwasborneoutthoughexamination

ofthefinaldictionarytermsincontext.Giventhelengthoftime

overwhichthesedatawerecollected,historicterminologiesforthe

conditionsstudiedwerealsoincluded,e.g.chronicobstructive

pul-monarydiseaseorheaves.Grasssicknessisveryrarelyencountered

initstruechronicform,butwasincludedhere(allformsthereof)

becauseitrepresentedadiseasewithawell-recognisedhigh

mor-talityrateandshortdiseasecourse.Grasssicknesstherebyactedas

aformofpositivecontrol,andallowedcomparisonwithconditions

oflowermortalityandmoreprotracteddiseasecourses.

Neopla-siaconstitutes a largenumber of differentdiagnosesin equine

medicine,whichmaybereachedfollowingaplethoraof

diagnos-tictests.Thisgroupwasnotsubdividedintospecifictumourtypes

duetosmallsamplesizes,butalsobecausethelevelofdiagnostic

accuracyreachedinfirstopinionpracticevarieswidely.Itcouldbe

arguedthatallthatisrequiredforahorsetobeeuthaniseddue

toneoplasiaisastrongsuspiciononthepartoftheveterinarian

(whichwouldlikelybeenteredinthemedicalrecords),irrespective

ofthefinalspecificdiagnosis(ifavailable).Diagnosis(presumptive

ordefinitive)ofneoplasiacarriesgravesignificanceforhorse

own-ers,andsimilartotheotherconditionsstudiedhere,couldinfluence

theirbehaviourtowardearlyeuthanasia.Overlap(co-occurrence)

interminology(cases)betweenneoplasia,PPIDandsarcoidswas

lowtomoderate.Betweensixandseventeenpercentofhorses

cate-gorisedunderoneofthesediseaseswassimultaneouslycategorised

underoneoftheothers(datanotshown),lendingconfidencetothe

decisiontomodeltheseconditionsseparately.

Deathinhorsesrarelyoccursnaturally,i.e.withouteuthanasia.

Anowner’sdecisiontoeuthanaseisveryoftencomplex,

involv-ingwelfare,management,financialandemotionalaspects.Itcan

besupposedthatdiagnosisofachronicorincurablediseaseoften

playsalargepartinthisdecision,butitwillrarelybetheonlyreason

foreuthanasia.Inaddition,itiscommonthattheprecisereasonfor

euthanasiaisnotnotedinmedicalrecords(aswasthenorminthe

datasetusedhere).Forthesereasons,all-causemortalitywasused

astheeventofinterestintheseanalyses,ascause-specific

infor-mationwasnotavailable.‘Survivaltimes’reportedheretherefore

reflecteithersurvivalfromdiagnosisuntilnaturaldeath(fromall

causes)oruntileuthanasia.Theuseofall-causemortalityis

neces-sarywhenusingdatathatoftenomitthecauseofdeath(orinclude

unreliableinformation),andisalsocommonlyemployedinhuman

andanimalepidemiologicstudies(Blacketal.,2002;Tillingetal.,

2002;Sallesetal.,2004;deMadronetal.,2011;Mattinetal.,2015).

Allofthechronicdiseasesstudiedwerefoundtosignificantly

affectsurvival.ItisofinterestthatthehazardratiosforEMSand

PPIDaresolarge,andthattheexpectedsurvivaltimesfollowing

diagnosisoftheseconditionsisshorterthanmostotherdiseases.

Equinemetabolicsyndromeisarelativelynewdiagnosisinequine

medicine,and,similartoitshumancounterpart,featuresmetabolic

derangements,obesityandinsulinresistance(McCueetal.,2015).

In horses,theconditionhasbeenlinkedwithanincreasedrisk

oflaminitis.AlthoughEMSisitselfnotlife-threatening,laminitis

certainlyisthroughthenecessityforeuthanasiawhenwelfareis

significantlycompromised.Similarly, PPIDisinextricably linked

withlaminitis,butdoesnotleadtonaturaldeathinthevast

major-ityofcases,thereforeitcouldbesupposedthatitseffectonlife

expectancyandsurvivalwouldbemodestoncelaminitiswastaken

intoaccount.Theseendocrinopathieshavebeenreportedto

com-monly co-occur,which is a feature of the present dataset(see

Table3)(Donaldsonetal.,2004;Karikoskietal.,2011;Irelandetal.,

2013).Despite takingbreed, ageat diagnosis,and co-occurring

diseasesintoaccount,theeffectonmortalityoftheseconditions

remainssubstantial.Farfrombeing‘incidental’featuresofolderage

inmanyhorses,thesediseasesareseriousdiagnosesthatwarrant

furtherresearch.

Co-andmultimorbiditywereafeatureofthesedata,with1.2%

ofhorsesdiagnosedwithmorethanonechronicdisease(maximum

6)overthestudyperiod.Laminitiswasafeatureof74%ofallhorses

withmultimorbidity.Laminitisisapainfulconditionofungulates

whereinflammationofthedermallaminaecanresultinseparation

oftheunderlyingsoft tissuestructures fromtheoverlying

pro-tective,suspensoryhoofwall.Thediseaseisacommonsequelto

manyinflammatoryconditions,andexhibitsbothacuteandchronic

phases.Giventhis‘finalcommonpathway’ofinflammatorydisease,

itisperhapsunsurprisingthatlaminitisissocommonlydiagnosed

alongsideotherchronicdiseases.PPIDwasafeaturein49%of

mul-timorbidanimals,andgiventhisdisease’sconsiderableeffecton

lifeexpectancyandhazardratios,itconstitutesaseriouschallenge

(8)

Table7

Medianlifeexpectancyfollowingcommencementoflifetimediseaseco-occurrence status(otherindependentvariablesheldattheirreferenceormedianlevels)ina conveniencesampleoftheveterinary-attendedUKgeneralhorsepopulation.

Lifetimedisease co-occurrencestatus

Medianlifeexpectancy inyearsfollowing commencementof status(95%CI)

Median(IQR)age (years)at commencementof lifetimeco-occurrence status

Nochronicdisease 19.8(19.4–20.6) 6(3–12) Onechronicdisease 11.5(10.2–13.5) 14(7–22) Twochronicdiseases 7.9(6.6–10.4) 18.5(11–25) Morethantwochronic

diseases

1.7(0.3–9.4) 23(18–29) CIconfidenceinterval.

Whencomparedwithhavingnoneofthechronicdiseases stud-ied,co-andmultimorbiditywerefoundtopose6.6and21.3times thehazardofdeath,respectively(Table6).Thisresultsuggeststhat

co-occurrenceofdiseasewithinanindividualposesadditionalrisk

tosurvival abovewhatmaybeexpectedsimplybytheproduct

oftheriskofbothdiseasesindividually.Oftheconditionsstudied

here,onlyneoplasiaandgrasssicknesscanleadtounassisteddeath

inthehorse,thusthedeathsofhorsessufferingfromtheremaining

diseasesislikelytobeviaeuthanasia.Asmentionedpreviously,the

decisionofwhentoeuthanasecanbecomplex,andinvolvesmore

thansimplythelevelsofpainordistressananimalisexperiencing.

Whenahorseisdiagnosedwithmultiplechronicdiseases,itis

pos-siblethateuthanasiaiscarriedoutmorereadilythanifthatanimal

onlyhadonedisease,duetocomplexityanddifficultyin

manag-ingandtreatingmultiplediseases,and/orduetoincreasedcostsof

treatment.Overthecourseofthisstudy,theapparentprevalence

ofmultimorbiditywasfoundtoincrease.Giventhesignificantand

increasingapparentprevalenceofmultimorbidityinthissample,

moreresearchintodiseaseclusteringandthemanagementof

co-occurringdiseasesiswarranted,inorderthatparsimoniousand

harmonioustreatmentregimenscanbefoundforcommondisease

groups,andunderlyingaetiologicalcommonalitiescanbebetter

understood(Table7).

Justasinhumans,theaveragelifespanofveterinary-attended

UK horses in this population appears to be increasing overall

(Wohlandetal.,2015).Thisresultisinagreementwith,butoffers

morevalidityandrobustnesstotheconclusionsofBrosnahanand

Paradis(2003),oftencitedinveterinaryliterature.TheDepartment

ofHealthoftheUKGovernmentoftenusescomparisonofthelife

expectancyatbirthbetweengroupstoassesswhetherapopulation

isageing.Thisrobustmethod,employedbytheOfficeforNational

Statistics,wasimplementedusingthecurrentdata.Themethod

allowstheuserto‘forecast’thelifeexpectancyofhorsesofdifferent

ages,evenwhenthedeathsinanagecohorthaveyettooccur.A

sig-nificantincreaseinlifeexpectancywasfoundbetweenhorsesborn

between2010and2012,comparedtothosebornbetween1995and

1997.Giventhelowoverallproportionofhorsesthatdiedoverthe

entirestudy(6.7%),itappearsthatdeathisoftennotrecorded.For

thisreason,survivaltimesreportedhereareusefulforcomparison

betweendiseases and comorbiditystatuses,however their

esti-matedmediansshouldbeinterpretedcautiously.Brosnahanand

Paradis(2003)reportedanincreaseintheproportionofhorsesof

twentyyearsoldormorebeingpresentedtoaUniversityveterinary

clinicbetween1989and1999(BrosnahanandParadis,2003).The

authorsconcludedthatthereasonforthisincreasewasunknown,

butlikelytobemultifactorial,andcitedgreaterequinelongevityas

alikelyphenomenon.Increasingwillingnessonthepartofowners

tofinanceveterinarycareofgeriatrichorseswasalsoposited.The

referralcasesusedinthestudybyBrosnahanandParadis(2003)are

unlikely,however,toberepresentativeofallveterinarian-attended

horses,norofthehorsepopulationasawhole.Itispossiblethat

theaveragelongevityofhorseswhoseownershad thefinancial

meanstoaccessreferralcarewasincreasingbetween1989and

1999,butthelongevityofhorsesreceivingfirst-opinioncareonly,

ornoveterinaryattention,wasnot.Thedatausedhereincludeda

largesampleoffirst-opinionmedicalrecords,overagreater

times-pan,andthuscannotbedirectlycomparedtothepreviousstudy.

Ofcourse,deathinthispopulationisveryoftenaresultof

euthana-sia,thus‘survivability’isnottheonlyvariableaffectinglongevity.

Financial,management,andwelfareaspectsallcontributetothe

decisiontoendananimal’slife.

The results of the current study taken together constitute

an important contributionto theunderstandingof equine

epi-demiology,andindicatethatveterinariansmayencountergreater

numbersofagedanimalsinfuture. Chronicdiseaseis prevalent

andofseriousconcernforequinewelfareandlongevity,andthe

apparentprevalenceofmultimorbidityappearstobeincreasing.

Furtherstudyofpatternsofmultimorbidityinthehorsemayhelp

tobetterunderstandaetiologicalcommonalitiesandtoformulate

appropriatemanagementstrategies.

Acknowledgments

Theauthorsaregratefultotheownersandveterinarianswho

contributeddatatothisstudy.C.W.isfundedbytheAnimalWelfare

Foundation(NormanHaywardFund20141JM’).

AppendixA. Supplementarydata

Supplementarydataassociatedwiththisarticlecanbefound,in

theonlineversion,athttp://dx.doi.org/10.1016/j.prevetmed.2016.

07.011.

References

Anholt,R.M.,Berezowski,J.,Jamal,I.,Ribble,C.,Stephen,C.,2014.Miningfree-text medicalrecordsforcompanionanimalentericsyndromesurveillance.Prev. Vet.Med.113,417–422.

AnimalHealthTrust(AHT),DepartmentforEnvironmentFoodandRuralAffairs (DEFRA),BritishEquineVeterinaryAssociation(BEVA),2015.EquineQuarterly DiseaseSurveillanceReport.AnimalHealthTrust,Newmarket,Suffolk,UK.

Black,W.C.,Haggstrom,D.A.,Welch,H.G.,2002.Re:all-causemortalityin randomizedtrialsofcancerscreening–response.J.Natl.CancerInst.94, 865–866.

Boden,L.A.,Parkin,T.D.H.,Yates,J.,Mellor,D.,Kao,R.R.,2013.Anonlinesurveyof horse-ownersinGreatBritain.BMCVet.Res.9.

BritishEquestrianTradeAssociation,2011.BETANationalEquestrianSurvey2011. Wetherby,WestYorkshire.

Brosnahan,M.M.,Paradis,M.R.,2003.Demographicandclinicalcharacteristicsof geriatrichorses:467cases(1989–1999).J.Am.Vet.Med.Assoc.223,93–98.

Chiang,C.L.,1968.IntroductiontoStochasticProcessesinBiostatistics.JohnWiley &Sons,Inc.,NewYork,London,Sydney,N.S.W.,Aust.

Chiang,C.L.,1978.LifeTableandMortalityAnalysis.WorldHealthOrganisation, Geneva.

Dardis,C.,2015.survMisc:MiscellaneousFunctionsforSurvivalData.Rpackage version0.4.6.

Dohoo,I.R.,2009.VeterinaryEpidemioloicResearch.VERInc.,Charlottetown.

Donaldson,M.T.,Jorgensen,A.J.R.,Beech,J.,2004.Evaluationofsuspectedpituitary parsintermediadysfunctioninhorseswithlaminitis.J.Am.Vet.Med.A224, 1123–1127.

Hotchkiss,J.W.,Reid,S.W.J.,Christley,R.M.,2007.Asurveyofhorseownersin GreatBritainregardinghorsesintheircare.Part2:riskfactorsforrecurrent airwayobstruction.EquineVet.J.39,301–308.

Ireland,J.L.,Wylie,C.E.,Collins,S.N.,Verheyen,K.L.P.,Newton,J.R.,2013.

Preventivehealthcareandowner-reporteddiseaseprevalenceofhorsesand poniesinGreatBritain.Res.Vet.Sci.95,418–424.

Islam,M.M.,Valderas,J.M.,Yen,L.,Dawda,P.,Jowsey,T.,McRae,I.S.,2014.

Multimorbidityandcomorbidityofchronicdiseasesamongthesenior Australians:prevalenceandpatterns.PLoSOne9.

Karikoski,N.P.,Horn,I.,McGowan,T.W.,McGowan,C.M.,2011.Theprevalenceof endocrinopathiclaminitisamonghorsespresentedforlaminitisata first-opinion/referralequinehospital.Domest.Anim.Endocrinol.41,111–117.

Lam,K.,Parkin,T.,Riggs,C.,Morgan,K.,2007.Useoffreetextclinicalrecordsin identifyingsyndromesandanalysinghealthdata.Vet.Rec.161,547–551.

(9)

Mattin,M.J.,Boswood,A.,Church,D.B.,McGreevy,P.D.,O’Neill,D.G.,Thomson,P.C., Brodbelt,D.C.,2015.Degenerativemitralvalvedisease:survivalofdogs attendingprimary-carepracticeinEngland.Prev.Vet.Med.122,436–442.

McCue,M.E.,Geor,R.J.,Schultz,N.,2015.Equinemetabolicsyndrome:acomplex diseaseinfluencedbygeneticsandtheenvironment.J.EquineVet.Sci.35, 367–375.

Mellor,D.J.,Love,S.,Gettinby,G.,Reid,S.W.J.,1999.Demographiccharacteristicsof theequinepopulationofnorthernBritain.Vet.Rec.145,299–304.

Pache,B.,Vollenweider,P.,Waeber,G.,Marques-Vidal,P.,2015.Prevalenceof measuredandreportedmultimorbidityinarepresentativesampleoftheSwiss population.BMCPublicHealth15.

RCoreDevelopmentTeam,2015.R:ALanguageandEnvironmentforStatistical Computing.RFoundationforStatisticalComputing,Vienna,Austria.

Salles,G.F.,Bloch,K.V.,Cardoso,C.R.L.,2004.Mortalityandpredictorsofmortality inacohortofBraziliantype2diabeticpatients.DiabetesCare27,1299–1305.

Slater,J.,2014.FindingsfromtheNationalEquineHealthSurvey,2013.Vet.Rec. 175,271–272.

Therneau,T.M.,Grambsch,P.M.,2000.ModelingSurvivalData:ExtendingtheCox Model.Springer-Verlag,NewYork.

Therneau,T.,2015.APackageforSurvivalAnalysisinS.

Tilling,K.,Sterne,J.A.C.,Szklo,M.,2002.Estimatingtheeffectofcardiovascularrisk factorsonall-causemortalityandincidenceofcoronaryheartdiseaseusing G-estimation–theAtherosclerosisRiskinCommunitiesStudy.Am.J. Epidemiol.155,710–718.

Wohland,P.,Rees,P.,Nazroo,J.,Jagger,C.,2015.Inequalitiesinhealthylife expectancybetweenethnicgroupsinEnglandandWalesin2001.Ethnic Health20,341–353.

Wylie,C.E.,Ireland,J.L.,Collins,S.N.,Verheyen,K.L.P.,Newton,J.R.,2013.

DemographicsandmanagementpracticesofhorsesandponiesinGreat Britain:across-sectionalstudy.Res.Vet.Sci.95,410–417.

deMadron,E.,King,J.N.,Strehlau,G.,White,R.V.,2011.Survivaland

echocardiographicdataindogswithcongestiveheartfailurecausedbymitral valvediseaseandtreatedbymultipledrugs:aretrospectivestudyof21cases. Can.Vet.J.Larevueveterinairecanadienne52,1219–1225.

References

Related documents

Production packer Mechanical firing head with automatic release Perforating guns Pup joint Ported disc assembly Production seal assembly Production tubing Figure 13 - Running In...

Turn on led_Red1 Turn on led_Green2 Wait 30 seconds Turn on led_Yellow1 Turn on led_Yellow2 Wait 3 seconds Turn on led_Red2 Turn on led_Green1 Wait 20 seconds Turn on led_Yellow1

... Most specimens were found in and around a boulder masonry fireplace on the south wall of the .French structure. This latter figure represents a major difference

The researcher met to the secondary school teachers in Thane District personally to collect relevant data. The purpose and nature of the study and the

Commercial aircraft programs inventory included the following amounts related to the 747 program: $448 of deferred production costs at December 31, 2011, net of previously

Survival analysis was carried out using Kaplan-Meier to obtain the mean and median survival time of the inmates and the hazard plot of the covariates while Cox regression

Contemporary art differs from modern and traditional art, thus it is relevant to ask if different methods should be used for teaching.. For example, which should be

The index showed significant associations with self-rated oral and general health status, as well as perceived dental treatment needs, satisfaction with mouth and prevalence of pain

•   Class average in the chemical kinetics was 8.2 points higher than the highest previous fall/spring course offering. •   Exit surveys of the students were positive, but

The forecast performance of DFM is evaluated in terms of the MAEs, by comparing it with SBVAR models, based on the FOSC and the RWA priors, besides the non-spatial models like the

However, the MD Regulations specify that the essential requirements (set out in paragraphs 8.7 and 13 of Annex I to Directive 93/42/EEC) regarding information on the package or

AIRWAYS ICPs: integrated care pathways for airway diseases; ARIA: Allergic Rhinitis and its Impact on Asthma; COPD: chronic obstructive pulmonary disease; DG: Directorate General;

Our proposed method advances the state-of-the-art by: (i) incorporating different knowledge sources in the identification of jargon (semantic information, subword in- formation,

of land area under organic management 36 Graph 6 Area under organic management – share per continent 37 Graph 7 Organic farms – share per continent 37 Graph 8 Arable

The literature review identifies characteristics of charitable organisations, markets and government and the emergent, hybridising nature of the dependence of charitable

Providing information on the level of contact required for the transfer of cellular material beneath female fingernails and on factors that affect DNA transfer

In fact, even if such approach appears as very powerful for detecting the presence of DoS attacks in the traffic, it still presents some important lacks: indeed, detecting a change

Keywords- behavior finance, Behavior biases, investment, Efficient market hypothesis, Financial decision making.. process,

The proposed indexing that uses Generalized Hash Tree (GHT) expedites projection and selection operations on encrypted medical XML records stored in WORM storage.. We implemented

As noted in the Literature Review, above, scholarship on the determinants of foreign direct investment (FDI) variously argue the influence of GDP growth, the openness of a

EHR Implement is collecting analyzing and comparing broad EHR‐Implement is collecting, analyzing and comparing broad scale Electronic Health Record implementations in European

This paper analyzed the lifespan of the Japanese information service industry by using Kaplan-Meier estimator and Cox proportional hazard regression model on a sample of 344