Risk and protective factors for falls on stairs in young children

26292_3909_full.pdf

Risk andpr otectivefactorsforfalls onstairsin

youngchildr en:multicentrecase -controls tudy

D Kendrick,

1

K Zou,

1

J Ablewhite,

1

M Watson,

2

C Coupland,

1 BKay, 3

A Hawkins,

4

R Reading

5 ▸Additional materialis published onlineonly .Tovie w please visitthe journalonline (http://dx.doi.org/10.1136/ archdischild-2015-308486). 1

Division ofPrimary Care,

School ofMedicine,

Nottingham,UK

2

School ofHealth Sciences,

Universityof Nottingham,

Queen's MedicalC entre,

Nottingham,UK

3

Emergency Department,

BristolChildr en's Hospital,

Bristol,UK

4

GreatNorthChildr en's

Hospital, NewcastleuponTyne

Hospitals NHSF oundation

Trust,Resear chUnitLevel2,

NewcastleuponT yne,UK

5

Jenny LindP aediatric

Department, Norfolkand

Norwich UniversityHospital,

Norfolk CommunityHealthand

CareNHS Trus t,Norwich,UK

Correspondenceto

ProfessorDenise Kendrick,

Division ofPrimary Care,

School ofMedicine, Floor13

TowerBuilding,University

Park,Nottingha mNG72RD,

UK; denise.kendrick@

nottingham.ac.uk

Received24F ebruary2015

Revised23 July2015

Accepted 6Augus t2015

Published OnlineFirs t

10 December2015

Tocite: Kendrick D,Zou K,

AblewhiteJ, et al.ArchDis

Child2016;101:909-916.

ABSTRACT

AimToinv estigateriskandprotective factors fors tair falls inchildr enaged<5y ears.

MethodsMulticentrecase-controls tudyathospita ls,

minor injuryunits andgener alpr acticesin andaround four UKs tudycentres.Cases werechildr enwit h medically attendedstair fallinjuries.C ontrols were matchedon age,se x,calendar timeandstudy centre. A total of610 casesand 2658contr olsparticipa ted.

ResultsCases'mostcommon injuries werebangs on

the head(66%), cuts/grazes notrequiring stitches (14%) and fractures(12%).Parents ofcases weresignificantly morelik elynottoha ve stair gates(adjustedOR (AOR)

2.50, 95%CI 1.90to 3.29;popula tiona ttributable

fraction(PAF) 21%)ortole av es tairga tesopen(AOR

3.09, 95%CI 2.39to 4.00;P AF24%) both compared

withhavi ngclosedsta irgates. Theyweremor elikelynotto havecar petedstairs(AOR1. 52,95%CI1.09t o2.10; PAF

5%)an dnotto havea land ingpart-wa yup theirstairs

(AOR1. 34,95%CI1. 08to1.6 5;P AF18%). The ywere morelik elytoconsiderth eir stairsunsafeto use(AOR1.4 6,

95%C I1.07t o1.99 ;PAF5% )ortobeinneed ofrepai r

(AOR1. 71,95%CI1. 16to2.5 0;P AF5%). ConclusionStructuralfactors includinghaving landings part-wayupthes tai rsandk eepingstairsi ngoodr epair wereasso ciatedwithreducedsta irfallin juryrisk.Fami ly factorsinclud inghavingstairgat es,notlea vinggatesop en andha vingstair carpetswereas sociatedwith reduced injuryrisk. Iftheseasso cia tionsareca usal,addressing thesefact orsinhousing policy androutinechi ldhealth promotioncouldr educestai rfallinjurie s.

INTRODUCTION

Fallsar etheleadingcause ofmedically attended

injury inchildr enaged<5y earsin most high- income countries. 1

In Englandand Wales, in2002,

the latestyearfor whichdetailedemergency department (ED)da taisavailable, fallsamong the childrenaged <5y earsr esultedinmore than

190 000ED attendances,

2 and in2012/2013 they accountedfor almost 20000hospitaladmissions in

England.

3

Fallsfr omstairsor stepscomprised18%

of EDa ttendancesforfalls 2 and 12%of admissions for falls. 3

While somefalls from stairsamongchil-

drenaged <5y earsar eassociatedwith objectssuch as babyw alkers,toysorpushchairs, most(88%) arenot andonly asmall proportion arisefr omchil- drenbeing dropped whilebeingcarriedon stairs. 4

A recentsy stematicoverviewfoundinterventions

providinghomesa fetyeduca tion,and/orhome safetyequipment wer eeffectiveinpromoting the use ofsa fetygateson stairsandsome evidencetha t

theyr educethenumberof familiesusing babywalkers.Theo vervie wfoundlittleevidencethatthese interventionsreduced injuryrates.

5

Our study

aimed toquantify riskand protectiv efa ctorsfor stairfalls amongchildr enaged <5years.

METHODS

The publishedpr otocolreportsfull detailsofthe

methods. 6

This studywas oneoffiveconcurr ent

case-controls tudies,eachr ecruitingchildrenwith one typeof injury( fallsfr omfurniture,falls on one level,stair falls,poisoning,scalds).

Study designandsetting

The studywas conductedinNHShospitals in

Nottingham, Bristol,Newcas tleuponTyne,

Norwich, Gateshead,Derby, LincolnandGrea t

Yarmouth,England. Caser ecruitmentcommenced

on 14June 2010and endedon 30September

2012. Controlrecruitmentcommenced atthesame

time ascase recruitment andendedwithin

4 monthsof caser ecruitment.

Whatis already knownonthis topic

▸Fallsar etheleadingcause ofmedical ly attendedinjury inpr eschool children,andfalls froms tairscomprise12% ofhospital admissionsand 18%of emergency department attendancesfor falls.

▸Homesa fetyeducationand equipmentprovisioncanincr easeuse ofstairsa fetyga tes,but thereislittle evidence that thisreducesinjuries.

Whatthis study adds

▸Familyfa ctorsincludinguseof stair gates, notleavingthemope nand havingas taircarpe twereassocia tedwithreducedrisk ofa stairfall

injury. ▸Structuralfacto rsincludinghavingalanding part-wayupthes tairsand keep ingstairsin goodr epairwere associatedwithredu ced injuryrisk.

▸If theseassocia tionsarecausal, addressingthese factorsinr outinechild healthpromot ionand housingpolicycould redu ces tairfallinjuries.

Kendrick D,et al.ArchDis Child2016;101:909-916. doi:10.1136/archdischild-2015-308486909

Original article

on August 11, 2023 by guest. Protected by copyright.http://adc.bmj.com/Arch Dis Child: first published as 10.1136/archdischild-2015-308486 on 1

0 December 2015. Downloaded from

Participants

Cases werechildrenaged 0-4 yearsattending EDs,minorinjury units oradmitted tohospital following afall onstairsin the child's home.Childr enwithintentionalor fatal injuryor living in residentialcare weree xcluded.Parents/carers ofpotentially eligible childrenwer einvitedtoparticipateduring theirmedical attendanceor bytelephone orpos twithin 72h ofattendance. Controlswer echildrenaged0-4 yearswithouta medically attendeds tairfall,recruited from thecase's generalpra ctice(or neighbouring practice).Weaimed torecruitan av erage offour controlsfor each casematchedon age(up to4monthsy ounger or 4months olderthan thecase), genderand calendartime (recruitedup to4 monthsof theda teof thecase injury).Study invitationsw eresentto10potentially eligiblecontr olsfor each case bymail from thepractice regis ter.Wheremore than10 controlparticipants metinclusion criteria,those withda tesof birth closesttotha tof theirmatchedcase wer echosen. To increasepo werandmakeef ficient useof recruited participants, controlparticipants from caseswithmore thanfour controls, controlsno longerma tchedto cases(eg,casehad subsequently been excluded)andcontrol participantsfr omtheotherfour ongoing case-controls tudieswere matched(onstudycentr e, age, genderand calendartime) tocases whichhad few erthan four controls. Participatingparents/car erscompletedage-specific(0-12, 13-

36 and≥37 months)ques tionnaires.Onereminderw asused for

non-responders.Those completingques tionnaires weregivena £5 giftv oucher.Othermethods,shown ina sy stematicr eview to increaseresponse rates,w ereused,includingpersonalised invitations,firstclass mailings,r emindersand inclusionofuni- versitylogos ons tudydocumenta tion. 7

Questionnairescollected

dataon exposur es,socio-demographicalandconfoundingvari- ables, injuriesand trea tmentreceived.

Sample size

Sample sizew asbasedon80% pow er, 5%signi ficance level, four controlspercase anda correla tionbetw eene xposuresin cases andma tchedcontrolsof 0.1.Todetect anOR of1.43,

496 casesand 1984contr olparticipants werer equired basedon

prevalenceof exposur esfrompreviouss tudies(babywalk eruse (36%), nosa fetygateson stairs(55%),not usingpla ypens (58%) andnot usings tationary activitycentres(76%)). 89

Exposures

The exposuresofinteres tw eresafetybehaviours, safetyequip- ment andhome hazardsr elating tostairs.Theseare describedin table 1, withr eportingperiods,response optionsand response categorisations.

Confounders

All analysesaccounted formatchingby ageand sexandadjus ted for distancefrom homeresidenceto hospital(ca tegorisedinto quintiles) 10 and IndexofMultiple Deprivation (IMD,linear term). 11

Distancefr omresidenceto hospitalandIMDw ere

included becausesome control participantscamefrom very dif- ferentneighbour hoodsthancaseparticipants andthe extr a matchedcontr olswere notmatchedonpra ctice.Dis tancefrom residenceto hospitalw asgr oupedintoquintiles( ≤2.0, 2.1-3.2,

3.3-4.7, 4.8-8.8, >8.8km). Directed acyclicgraphs werecon-

structedfor each exposureto identifytheminimalsufficient adjustmentset ofconfounders that analyses neededtoadjust for. 12 Potentialconfounders entered intodirecteda cyclicgr aphs

werefirstchild (y es/no);overcro wding(yes/no);ethnicgroup(white/other); singleadult household( yes/no); theHospital

Anxiety andDepr essionScale(HADS,linear term)

13 ; Parenting

Daily HasslesScale (PDH,linear term)

14 15 ; childbeha viour questionnairescore (linearterm) 16-18 ; hoursof out-of-home child careperw eek(linear term)andchild's abilityto open safetyga te(likely/notlik ely).Someexposur eswere alsoconsid- eredas potentialconfounders forother exposur esincluding use of playpen,teaching safetyruleson stairs,stair gates andthe composite stairsafety variabledescribedintable 1. Analysesfor eache xposurewereadjus tedforthoseconfoundersidentified in the directedacyclic graphsasbeing intheminimalsuf ficient adjustmentset (listed intable 3).

Statisticalanalysis

Associationsbetw eenexposures andstairfallswer ees timated using ORsand 95%CIs, usingconditional logistic regr ession adjustedfor confoundersas describedabo ve. Thelinearity of relationshipsbetween continuousconfoundersandcase/contr ol statuswas testedbyadding higher-ordertermsto regr ession models, withca tegorisationwherethere wassignificant non- linearity.Inter actiontermswere addedto regressionmodelsto exploredifferential effectsbychildage, gender, ethnicgr oup, single parenthood,non-owner -occupiedhousingandunemploy- ment. Aninter actionbetweenuseof babywalkers anduse of safetyga tesonstairs was alsoexamined.Signi ficance ofinter ac- tions wasassessedwith likelihood ra tiotests( p<0.01)and stratified ORpr esentedwheresigni ficant interactionswere found. Thepopula tionattributablefr action(PAF) percentage wascalcula tedforexposur eswith statisticallysignificantly raised adjustedORs (AORs) usingapublishedformula. 19 Forthe HADS,single missingitem valuesfor each subscale wereimputedusing themean ofthe remaining sixitems. Subscale scoreswer enotcomputedwhenmore thanone item wasmissing. 20

The sameappr oachwasusedfor missingvalues

of PDH,since we wereunable tofind specific guidanceon this. The mainanaly seswere completecase(CC)analysesincluding single imputedvalues forHADS andPDH. Casesand controls with responsesof'not applicable'wereex cludedfromanalyses whereappr opriate.Sensitivityanalysesw ere performedusing multiple imputation(MI),with case/control sta tus,studycentr e, age andse xofchild,IMD, distance from hospital,socio- demographicalchar acteristicsandallexposure andconfounding variables includedin theimputa tionmodel. Thisincluded imputationof HADSand PDHscor esfor casesand controls who hadmor ethanoneitem missingon thesescales. Tw enty imputed datasetswer ecreatedandthe resultswerecombined using Rubin 's rules. 21

Ethical approval

The studywas approved byNottinghamshireresearchethics committee. Informedconsent was assumedthroughr eturnof completed studyquestionnair es.

RESULTS

A totalof 610cases and2658 controls participated assho wnin figure1 . Thirtythr eepercentof parents/carersof casesand

29% ofcontr olsagreedto participate.Childparticipants and

non-participants weresimilarinterms ofse x(50% vs53% male), buta higherpr oportionof participantswere aged0 -

12 monthsthan non-participants(19% vs12%). Themean

number ofcontr olspercasew as4.36. Themedian timefrom dateof injuryto date ofques tionnairecompletionfor casesw as

11 days(IQR7-21). Mostcasessus tainedsingle injuries(85%),

mostcommonly bangson thehead (66%),cuts/gr azesnot

910Kendrick D,et al.ArchDis Child2016;101:909-916. doi:10.1136/archdischild-2015-308486

Original article

on August 11, 2023 by guest. Protected by copyright.http://adc.bmj.com/Arch Dis Child: first published as 10.1136/archdischild-2015-308486 on 1

0 December 2015. Downloaded from

requirings titches(14%)andbr oken bones(12%). Mostcases (64%) wereseenande xaminedbut didnot requiretr eatment,

25% weretreated inEDand5%wer eadmitted tohospital.

Table2 showstha tcomparedwith controls,casesw ere less likelyto live inahouseholdwith twoor more adultsin paid work (50%vs 59%).Cases alsoliv edin more deprivedareas (median IMDscor e18.7vs15.2) andw ere more likely tolive in singleadult households(15% vs11%), receiving sta tebene fits (41% vs32%), innon-o wner-occupied housing(40%vs32%), in householdswithout acar (15%vs 10%)and withmothers who hadtheir firstchild underthe ageof 20y ears(19% vs 9%).

Table3 showsfr equencyofexposur esand ORsfortheCC

and MIanaly ses.Compared withcontrols,caseparents wer esignificantly morelikely toleave stair gatesopen(AOR3.09,

95% CI2.39 to4.00) orto notha ve stair gates (AOR2.50,

95% CI1.90 to3.29), tonot hav ecarpeted stairs (AOR1.51,

95% CI1.09 to2.10) orto notha ve alanding part-wa yup

their stairs(AOR 1.34,95%CI1.08 to1.65). They wer emor e likelyto considertheir stairs notsa fetouse(A OR1.46, 95%CI

1.07 to1.99) orin needof repair (AOR 1.71,95% CI1.16to

2.50). Casehouseholds wer esignificantly lesslik elytohav etrip-

ping hazardson theirs tairs(A OR0.77,95%CI 0.62to0.97) or notha vehandrailsonall stairs(AOR 0.83,95% CI0.75 to

0.93). TheP AFpercentager angedfrom5% forvariousstair

featuresto18% fornot having alanding part-wa yuptheir stairs,to 21%for notha vinga stair gateandto 24%for leaving a safetygate openonstairs. Table1 Exposuresand reporting periods,responseoptions andca tegorisationofr esponses

Exposuresr eported24hprior toinjury (cases)or 24h priorto completingques tionnaire(co ntrols) Responseoption

Safetyga tes

a. Usedanywher einthehom eYes/no b. Usedon stairs Yes/no g

Groupedinto:

closed stairgate/ga te left open/noga tec. Lefts tairgateopen

Yes/no

Stair features

a. Landingpart-w ayupYes/no b. Spiralstairs orwindingstair cases Yes/no c. Handrailsons tairsOn alls tairs/onsomestairs/no d. Banister/railingonstairs On alls tairs/onsomestairs/no

Groupedinto: onall stairs vsother responses

e. Banister/railing:widthofbigges tgap Inches

Groupedinto: tertiles

f. Stairco veringCarpet/wood/metal/concrete/lino/vinyl/don't know/other

Groupedinto: carpetvs otherr esponses

g. Stairsar etoosteep Agree/neitheragr eeordisagree/disagr ee h. Stairsar etoonarro wAgree/neitheragr eeordisagree/disagr ee i. Stairsar epoorlylitAgree/neitheragr eeordisagree/disagr ee j. Stepsar einneedof repair Agree/neitheragr eeordisagree/disagr ee k. Banister/handrailisinneed ofr epairAgree/neitheragr eeordisagree/disagr ee l. Stairco veringisinneedof repair Agree/neitheragr eeordisagree/disagr ee

m. Stairsar esafeto useAgree/neitheragr eeordisagree/disagr eeGroupedinto: agree vsotherresponses A compositestair safetyvariable (foruseas a

confounder inanaly ses)includeditems (g)to (m) groupedas:

No saferesponses: agreetoall ofitems(g) to

(l)and disagree to(m)

All saferesponses: disagreetoall items(g)to

(l)and agree toitem(m)

Some saferespo nses:allothercombinationsof

responses

Use ofbaby walk ers(ages0-36 monthsonly) Yes/no

Use ofplay pensortra vel cotswhilechildawake (ages0 -36 monthsonly)Y es/no Use ofs tationaryactivitycentres(ages 0-36 monthsonly) Yes/no

Exposuresr eportedfor1w eekprior toinjury (cases)or1w eekprior tocompletin gques tionnaire (controls) Response option

Trippinghazards ons tairsEveryda y/mostdays/someda ys/never/doesnot apply

Groupedinto: at leastsomeda ysvsnev er

Exposuresev erreportedprior toinjury(cases)or priorto completingque stionnair e(contr ols)R esponseoption

Taughtchild safety rulesorinstructions about

a. Howtobeha ve whengoingdownthes tairsY es/no b. Carryingbig/lots ofthings whilegoing down thes tairsY es/no c. Leavingthingson stairs Yes/no Kendrick D,et al.ArchDis Child2016;101:909-916. doi:10.1136/archdischild-2015-308486911

Original article

on August 11, 2023 by guest. Protected by copyright.http://adc.bmj.com/Arch Dis Child: first published as 10.1136/archdischild-2015-308486 on 1

0 December 2015. Downloaded from

AORsfr omtheMIanaly sesdiffer edonly by>10%from the CC analysesforfour exposur es(not havingcarpetedstairs(11% higher inMI thanCC analysis), stair coveringin needofrepair (11% higherin MIthan CCanaly sis),banis terwidth ≥3.75 inches(20% higherin MIthan CCanaly sis)and useof stationaryactivity centre(11%lo werinMI thanCC analysis)). Therew ereseveralsigni ficant interactions(seeonlinesupple- mentary tableS1). Compar edwithhavingas tairga tethatw as

keptclosed, leaving stairgates openincreasedthe oddsofas tairfall injuryin thoseaged 0-36 months,with aparticularly high

odds ofinjury amongchildr enaged 0-12 months.Not having a stairga teincreasedthe oddsofas tairfall injuryin allage groups,again withhigher oddsin younger children, butthe dif- ferencebetw eenagegroups was lessmarkedthan forleaving gatesopen. Ther elationship betweenstairgate useandstair fall injuries alsodiffer edbetweenfamilies whousedanddid notuse baby walkers.Leavings tairgatesopen ornothaving astairga te increasedthe oddsof injury(compar edwith having aclosed Table2 Socio-demographicalchar acteristicsofcasesandcontrols

CharacteristicsCases

n=610 (%)Controlsn=2658 (%)

Study centre

Nottingham252 (41.3)1055 (39.7)

Bristol178 (29.2)796 (29.9)

Norwich97 (15.9)457 (17.2)

Newcastle83 (13.6)350 (13.2)

Median childage (ye ars)(IQR)*2.0(1.2

-2.9) 2.0(1.3 -3.1)

Age inmonths*

0-12113 (18.5)315 (11.9)

13-36362 (59.3)1694 (63.7)

37-62135 (22.1)649 (24.4)

Male299 (49.0)1320 (49.7)

Ethnic groupwhite547 (91.5)[12] 2371(91.0) [52]

Number ofchildr enaged<5 yearsinfamily †[8] [44]

07 (1.2)28 (1.1)

1358 (59.5)1566 (59.9)

2212 (35.2)911 (34.9)

≥325 (4.1)109 (4.2)

Firstchild 242 (43.3)[51] 1067(44.5) [260]

Maternalage <20a tbirth offirstchild ‡100 (18.5)[7] 219(9.1) [15]

Single adulthousehol d87(14.6)[15] 272(10.5) [76]

Median weeklyhoursout- of-homechild care(IQR)13.5 (1.0 -22.5) [43]15 (3.0-24.0) [165]

Adults inpaid work[16][56]

088 (14.8)284 (10.9)

1209 (35.2)784 (30.1)

≥2297 (50.0)1534 (59.0) Receivessta tebenefits241 (40.9)[21] 838 (32.4)[68] Overcrowding(>1personperr oom)52 (9.1)[40] 187 (7.5)[152 ] Non-owner-occupiedhousing241 (40.4)[14] 836 (32.2)[65]

Household hasno car88 (14.7)[12] 254 (9.7)[50]

Median IndexofMultiple Deprivation score (IQR)§18.7 (10.1-32.7)15.2 (9.0-27.1) [35] Median kilometresfrom hospital(IQR)3.4 (2.2-5.4)3.9 (2.4-7.6) [34] Mean ChildBeha viourQuestionnaire score(SD)§4.7 (0.9)[43] 4.6 (0.9)[293] Long-term healthcondition 63 (10.4)[6] 202 (7.6)[19] Median ChildHealth VisualAnalogue Scale(r ange0 -10) (IQR)§9.9 (9.0-10.0) [9]9.7 (8.4-10.0) [19] Median Health-RelatedQualityofLife inchildren≥2 years(PedsQL) (IQR)§¶(n=303) [6]

91.7 (83.3-97.6)(n=1342)[18]

89.3 (82.1-94.0)

Parentalassessmentof child's abilityto opensa fetyga te[21][111]

Not likely433 (73.5)1937 (76.1)

Veryor quitelik ely156 (26.5)610 (24.0)

Median ParentingDailyHasslesT asksscale (IQR)§,**14.0 (10.0-18.0) [61]14.0 (11.0-18.0) [152] Mean HospitalAnxiety andDepr essionScale (SD)§,**10.4(6.2)[14] 10.7(5.9) [36] [ ]missing values. *Age whenques tionnairecompleted.

†Some familiesreported zerobecausechild renwer eaged <5yearsattim eofinjuryorat timeof sendingcont rolquestionnaire, butaged>5y earswhenques tionnairew ascomplet ed.

‡Only applicablewhere motherscompleted questionnair e.

§A higherIndex ofMultiple Depriva tionscore indicatesgreaterdepriva tion.A higherChildBeha viourQuestionnaire scoreindicatesmo re activ eandmor eintensebehaviour.Ahigher

ParentingDaily HasslesT asksscale scoreindicates morehassle. AhigherHospitalAn xietyandDepr essionScale score indica tesgr eater symptomsof anxiety/depr ession.Ahigherscore

of ChildHealth VisualAnalogu eScale indicatesbetter health.A higherPedsQLscoreindicates betterqua lityof life.

¶Missing valuesr efertothosewith ≥50% itemson anyscale missing. **Missing valuesr efertothosewith more thanone itemmissing.

PedsQL,theP ediatric QualityofLifeInventory .

912Kendrick D,et al.ArchDis Child2016;101:909-916. doi:10.1136/archdischild-2015-308486

Original article

on August 11, 2023 by guest. Protected by copyright.http://adc.bmj.com/Arch Dis Child: first published as 10.1136/archdischild-2015-308486 on 1

0 December 2015. Downloaded from

Table3 Frequencyof exposur esincaseandcontrol participants,adjus tedORs from completecaseandmultiple imputation analysesand

populationa ttributablefraction (PAF)percentage

ExposuresCases

n=610Controlsn=2658AdjustedOR (95%CI) PAF (%)Confounders adjusted for§* Safetyga tesanywherein house†[12] [124]n=1921-HADS, PDH,firs tchild,stair safety , hours out-of-homechild care Used 465(76.3) 2013(79.4) 1[r eference]

1.22 (0.92to 1.62)Did notuse 142(23.8) 521(20.6)

Exposuresonly forhouseholds with

stairsCases Controlsn=598 n=2476 [6] [7]

Stair gate†[13] [40]n=2401

24

21Child's abilityto opensa fetyga te,taughtchild

rules aboutgoing down thestairs,c arryingthings downthe stairs, leavingthingson stairs,stair safety Gateclosed 174(29.7) 1245(51.1) 1[r eference] Gateleft open210 (35.9)555 (22.8)3.09 (2.39to 4.00)

No gate201(34.4) 636(26.1) 2.50(1.90 to3.29)

Carpeted stairs†[8] [28]n=2394 5HADS, PDH,s tairsa fety

Had 507(85.9) 2248(91.8) 1[r eference]

Did notha ve83(14.1)200(8. 2)1.52 (1.09to 2.10)

Landing part-wayupthes tairs†[5] [28]n=2766

18Stair safety

Had 180(30.4) 892(36.4) 1[r eference]

Did notha ve413(69.6)1556(63.6) 1.34(1.08 to1.65) Spiralor windings tairs†[7] [30]n=2757 -Stair safety

Did notha ve495(83.8)2044(83.6) 1[r eference]

Had 96(16.2) 402(16.4) 0.97(0.75 to1.27)

Trippinghazards ons tairs‡[18] [51]n=2367 -HADS, PDH,s tairsafety

Did notha ve397(68.4)1493(61.6) 1[r eference]

Had 183(31.5) 932(38.4) 0.77(0.62 to0.97)

Stairs toos teep

$ †[18] [80]n=2744 -Stair safety Other responses362(62.4) 1653(69.0) 1[r eference]

Agree 218(37.6) 743(31.0) 1.21(0.94 to1.56)

Stairs toonarr ow

$ †[23] [98]n=2742 -Stair safety Other responses421(73.2) 1894(79.7) 1[r eference]

Agree 154(26.8) 484(20.4) 1.28(0.96 to1.70)

Stairs poorlylit

$ †[26] [94]n=2380 -HADS, PDH,s tairsafety Other responses469(82.0) 2053(86.2) 1[r eference]

Agree 103(18.0) 329(13.8) 1.32(0.97 to1.79)

Steps inneed ofr epair

$ †[25] [96]n=2378 5HADS, PDH,s tairsa fety Other responses506(88.3) 2233(93.8) 1[r eference]

Agree 67(11.7) 147(6. 2)1.71(1.16 to2.50)

Banister/handrailons tairsin needof

repair $ †[32] [98]n=2377 -HADS, PDH,s tairsafety Other responses498(88.0) 2175(91.5) 1[r eference]

Agree 68(12.0) 203(8. 5)1.32 (0.92to1.88)

Stair coveringinneedof repair

$ †[26] [96]n=2378 HADS,PDH, stair safety Other responses501(87.6) 2205(92.6) 1[r eference]

Agree 71(12.4) 175(7. 4)1.41 (0.99to2.03)

Stairs safeto use

$ †[10] [25]n=2391 5HADS, PDH,s tairsa fety Other responses487(82.8) 2180(88.9) 1[r eference]

Disagree101 (17.2)271 (11.1)1.46 (1.07to 1.99)

Handrailson alls tairs

$ †[1] [20]n=2416 -HADS, PDH,s tairsafety

Had 382(64.0) 1393(56.7) 1[r eference]

Did notha ve215(36.0)1063(43.3) 0.69(0.56 to0.86)

Banistersor railings onallstairs

$ †[22] [60]n=2301 -HADS, PDH,s tairsafety

Had 424(73.6) 1930(79.9) 1[r eference]

Did notha ve152(26.4)486(20.1) 1.27(0.99 to1.63)

Rules aboutgoing down thestairs[20] [70]n=1840 -HADS, PDH,firs tchild,child's ability

to opensa fetygate,s tairgate,s tairsafetyHad taughtchild405 (70.1)1782 (74.1)1 [refer ence]Had nottaught child173 (29.9)624 (25.9)1.36 (0.92to 2.02)Rules aboutcarrying thingswhile going

downthe stairs [20] [68]n=1840 -HADS, PDH,firs tchild,child's abilityto opensa fetyga te, stairga te,stairsa fety Had taughtchild287 (49.7)1274 (52.9)1 [refer ence] Had nottaught child291 (50.4)1134 (47.1)1.21 (0.83to 1.75)

Continued

Kendrick D,et al.ArchDis Child2016;101:909-916. doi:10.1136/archdischild-2015-308486913

Original article

on August 11, 2023 by guest. Protected by copyright.http://adc.bmj.com/Arch Dis Child: first published as 10.1136/archdischild-2015-308486 on 1

0 December 2015. Downloaded from

stairga te)amongwalk erusers andnon-users,buttheoddsof injury wereparticularlyhighin walk erusers wholeft gates open. Therewer esignificant interactionsbetweenthe number of adultsin paidwork andtea chingrules about(1) carrying things whilegoing down thestairsand (2)lea vingthingson stairs.Not teaching childreneitherrule reducedtheodds ofa stairfall injuryin familieswher enone ofthe adultswere employedbut notin familieswith employed adults.Ther ew ere also significant interactionsbetweenthe numberofadultsin the household andtwo exposur es.Therewasa reducedoddsof a stairfall injuryin singleadult householdsnot teaching rules about leavingthingson stairs butnot inhouseholdswithtwo or moreadults. There wasanincr easedoddsofa stair fallinjuryin single adulthouseholds withoutcarpeted stairs butnot in households withtwo ormor eadults.

DISCUSSION

A rangeoffa ctors,mos tofwhichw eremodifiable, increased the oddsof stair fallsinchildren aged0 -4 years.Thisincluded not usingsa fetygateson stairsorlea vingga tesopen,particu- larly infamilies alsousing babyw alkers, notha vingcarpeton stairs,not having alandingpart-wa yup thes tairs,havingstairs thatw ereinneedofr epairor having stairs thatparents per- ceivedto beunsa fe.The PAFranged from 5%to24%for these factorsindividually ,but45%of stair fallinjuries couldbe pre- ventedby having stairgates andkeepingthem closedassuming our associationsare causal.

Strengthsand limitations

This largecase -controls tudytookplace inEnglish NHShospi-

tals andincluded bothurban andrur alar easwith variedlevelsof socioeconomicdepriva tion(rangingfr om10%ofpopula tion

living inthe 20%mos tdepriv edareasin EnglandforNorfolk to 52%for Nottingham). 22

Weadjus tedforawide range of

potential confoundingfa ctorsselectedusingdir ecteda cyclic graphs.Analy sesusingmultiplyimputed data rev ealedbr oadly similarfindings toCC analyses.

Our participationra tewaslow(33% ofcasesand29% of

controls).P articipationratesw eresimilarbysex ofchild,buta higher proportionofparticipants wer eaged 0-12 monthsthan non-participants. Wecouldnot collecte xposure data fromnon- participants andhence thee xtentto whichselectionbiasma y haveoccurred isunknown.Self -reported exposuresma yhave been subjectto recall orsocialdesirability bias.Our casesw ere moredisadvantaged thancontr ols,and someexposures may havebeenassocia tedwith disadvantage.Althoughwe adjusted for arealevel deprivation,somer esidualconfoundingmay hav e remained.F ailuretofind significant associationsfore xposures whose prevalencewas outsidetherange usedin oursample size calculation( table 3; spiral/windingstairs, narrows tairs,poorly lit stairs,banister/handr ailinneedofrepair ,use ofpla ypenor stationaryactivity centre)may beduetoinsuf ficient power. Wefound sever al'counter-intuitive'findings. Childrenliving in houseswithout handrails onallstairs andthose withtripping hazards ons tairshadlow erodds ofinjury.Itispossible that if parentsper ceivestairstobe unsafethey may res trictaccessto the stairsor supervisechildr endiffer ently.Ourfindings regarding the increasedoddsof afall associated withtea chingchildr en safetyrules infamilies withunemploy edor singlepar entsmay reflect thepoor erqualityandmor ehazardous housingin which such familiesma ylive, 23
and despiteadjus tingforar angeof con- founders, thisma ynothav ebeen takenfullyintoaccount inour

Table3 Continued

ExposuresCases

n=610Controlsn=2658AdjustedOR (95%CI) PAF (%)Confounders adjusted for§*

Rules aboutlea vingthingson stairs[22][64] n=1838-HADS, PDH,firs tchild,child's abilityto opensa fetyga te,

stairga te,stairsa fetyHad taughtchild256 (44.4)1073 (44.5)1 [refer ence] Had nottaught child320 (55.6)1339 (55.5)0.85 (0.60to 1.22)

Exposureonly forhouseholds withs tairs

and banistersCases(n=424)Controls(n=1930) Banisterwidth (inches)[190] [803]n=627-Stair safety

0-2.594 (40.2)400 (35.5)1 [refer ence]

2.5-3.7567 (28.6)363 (32.2)0.83 (0.53to 1.29)

≥3.7573 (31.2)364 (32.3)0.75 (0.48to 1.18)

Exposuresonly forchildr enaged

0-36 monthsCases(n= 475)Controls(n=2009)

Baby walker[14] [32]n=1620-HADS, PDH,firs tchild,hoursout-of-home childcar e

Did notuse 326 (70.7)1302 (65.9)1 [refer ence]

Used135 (29.3)675 (34.1)0.83 (0.63to 1.10)

Playpenor tra velcot[14] [30]n=1615-HADS, PDH,used babyw alker, firstchild,hours out-of-homechild care Used77 (16.7)334 (16.9)1 [refer ence]

Did notuse 384 (83.3)1645 (83.1)1.07 (0.75to 1.53)

Stationarya ctivitycentre [16] [33]n=1611-HADS, PDH,used babyw alker, firstchild,hours out-of-homechild care Used111 (24.2)490 (24.8)1 [refer ence]

Did notuse 348 (75.8)1486 (75.2)1.08 (0.80to 1.46) [ ]missi ngvalues.

Stair safetyisa compositevariable combining responses toques tionsmarkedwith $andgrouped asall 'safe'responses,some 'safe'responsesandno 'safe'responses.Where the

exposurevariable isa measureof stair safety, thisvariable isexcludedfr omthecomposite stair safety measure usedasaconfoun derinadjus tedanal yses.

*Conditionallogistic regressione xcludesobservationswhere allcasesandtheir matchedcontrolsha vethesame exposure.

†In thepas t24h. ‡In thepas tweek.

§All modelsw ereadjustedfortheIndex ofMultiple Depriva tionand distance from thehospital pluslistedconfounders.

HADS, HospitalAnxiety andDepressio nScale; PAF,popu lation attributablefraction;PDH,P arentingDailyHasslesScale.

914Kendrick D,et al.ArchDis Child2016;101:909-916. doi:10.1136/archdischild-2015-308486

Original article

on August 11, 2023 by guest. Protected by copyright.http://adc.bmj.com/Arch Dis Child: first published as 10.1136/archdischild-2015-308486 on 1

0 December 2015. Downloaded from

analyses.As we exploredassocia tionsbetweenmanyexposur es and falls,some significantfindings mayrepr esenttype1errors.

Comparisonswithe xisting literature

Weha vefoundonlyonesmall Austr aliancase -controls tudyof infants withhead orfa ceinjuries withwhichtocompar eour findings. Thes tudyexaminedassocia tionsbetweenuse ofsafety gatesand fallsin familiesusing babyw alkers, andfound that households usingbaby walk erswithoutstairguardsor barriers had a3.5-fold increased riskofchildhead injurythan those using guardsor barriers(OR 3.53,95% CI1.21 to10.30). 24
Our studyextends thefindings ofthe Austr alianstudyby showingtha ttheoddsof as tairfall wer eparticularly highin families whoused babyw alkers andleftstairgates open.This maybe partlye xplainedby riskcompensationif familieswho use safetygates usewalkers upstairs moreoften,feel'safer' using walkersupstairsor forgettoclosega tesmor eoften.

CONCLUSION

If theassocia tionswefound arecausal,use ofsa fetygateson stairs,not leaving safetygates open,particularlyinfamilies also using babyw alkers,usingcarpetons tairs,k eepings tairsin good repairand having alandingpart-wa yup thes tairscouldindi- vidually preventbetween5% and24%ofinjuries from fallson stairsand, iffamilies hads tairga tesand keptthemclosed, 45% of injuriescould bepr evented. Thisadvicecouldbeincludedin child healthpr omotionprogrammes, personalchildhealth records,home safety assessmentsandotherchild healthcon- tacts.F utureresearchis neededtoexploreassocia tionsbetw een some stairchara cteristics,useofplaypensands tationary activity centresand injuryoccurr ence. AcknowledgementsThe authorswish tothank thepar entswho participated in the study.Theywould alsoliketo thankthe PrincipalInv estigators,liaison health visitors, researchnursesandother sta fffr omthe emergencydepartmentsandminor injury unitswho assisted inrecruitingparticipants from NottinghamUniversity Hospitals NHST rust,DerbyHospitalsNHSF oundation Trus t,United Lincolnshire Hospitals NHST rust,NorfolkandNorwichUniv ersityHospitals NHSF oundation Trust,JamesP agetUniv ersityHospitalsNHSF oundationTrus t,Univ ersityHospitals BristolNHS Founda tionTrust,NorthBris tolHealthcareTrust, Newcas tleuponTyne Hospitals NHSF oundationTrust,Ga tesheadNHSFoundationT rustandNorthumbr ia HealthcareNHSFounda tionT rust.Theygratefull yacknowledgethesu pportprovi dedfor recruitmentbythePrima ry CareR esearchNetworks forEastM idlandsandSouth Yorkshire,Leicest ershire,NorthamptonshireandRu tland,EastofEngland,Northernand Yorkshireandfro mSout hWestandT rent,Norfol kandSuffolk, North umberlandTyne andW earandWes tern ComprehensiveLocal ResearchNetworks.Written permissio n hasb eenobtain edandtheauthors thankP ennyBen ford,ClareTimb lin,Philip Miller, JaneSte wart,PersephoneWy nnandBenYoungfromthe Univers ityofN ottingh am; GosiaMajsak -Newman,LisaMcDaid,Clare Fernsand NathalieHorncast lefromthe NorfolkandNo rwichUn iversityHospi talsNHSFoundationTrus t;ToityDeave,T rudy Goodenough,PilarMuno zandBenit aLaird-Hop kinsf romtheUnivers ityoftheWestof England,PaulH indmarch, EmmaDavisonandLauraSi mmsfro mtheGreat North Children's Hospital,Newcas tleuponTynewhohelpedw ithrec ruitment ,data collection, prepareddataf oranalysisorc omment edondraftsof pap ers.Theywouldliket o acknowledgethefollo wingPrinci palInvestigatorsw hocontributedin obtaining funding,studyd esign,projec tmanagementint heircentresandinin terpretinganaly ses andc ommentedondrafts ofpaper s:ElizabethTow ner (Universityofth eWestof England),ElaineM cColl(Newc astleUniversity ),AlexJSuttonandNicola Cooper (UniversityofLeicest er) andFrankCoffey(Nott inghamUn iversit yHospitalsNHST rust). Theyare alsover ygrat efultoRo seClacy,layresea rch adviser,w hoattendedpr oject managementmeeting s,helpeddraftand pilotstudyd ocumentation,adv isedon recruitmentstrate giesandcommentedondrafts ofth epaper. ContributorsDK hadthe originalidea forthe study ,designed andsupervised the study,analysed andinterpretedda ta,dr aftedsectionsof themanuscript,edited and approvedthesubmittedv ersion.KZ analysed andinterpretedda ta,dr aftedsections of themanuscript, criticallyr eview edthemanuscriptandapprov edthe submitted version.JA collectedand interpreted data, draftedsections ofthemanuscript, critically reviewedthemanuscriptandappr ov edthe submittedv ersion.MWdesigned the study,interpretedda ta,drafted sectionsofthemanuscript, criticallyreviewedthe Figure1 Flowof casesand controls through study(stair fall). Kendrick D,et al.ArchDis Child2016;101:909-916. doi:10.1136/archdischild-2015-308486915

Original article

on August 11, 2023 by guest. Protected by copyright.http://adc.bmj.com/Arch Dis Child: first published as 10.1136/archdischild-2015-308486 on 1

0 December 2015. Downloaded from

manuscript andappr ovedthesubmittedversion.CC designedthe study ,analysed and interpreteddata, criticallyreview edthe manuscriptandapprovedthesubmitted version.BK andAH collectedand interpreted data, criticallyr eviewedthe manuscript and approvedthesubmittedversion. RRdesigned andsupervised thestudy, interpretedda ta,drafted sectionsofthemanuscript,critically revie wed the manuscript andappr ovedthesubmittedversion. FundingThis paperpr esentsindependentresear chfunded bytheNationalInstitute for HealthR esearch(NIHR)underitsPr ogramme Grants forApplied Research Programme(RP-PG-0407-10231). Theviewse xpressed inthisarticlear ethoseof the authorsand notnecessarily thoseof theNHS, theNIHR orthe Departmentof

Health.

Competinginteres tsNone declared.

Ethics approvalNorth Nottinghamshireresearch ethicscommittee. Provenanceandpeerr eview Not commissioned;externally peerreview ed. Open AccessThis isan OpenAccess articledis tributedin accordance withthe terms ofthe Crea tiveCommonsAttribution(CCBY4.0) license,whichpermits others todis tribute,remix,adapt andbuilduponthis work,for commercial use, providedtheoriginal workis properly cited.See: http://crea tivecommons.org/ licenses/by/4.0/

REFERENCES

1 PedenM,Oy egbiteK, Ozanne-SmithJ,et al.Worldrep ortonchildinjury

prevention. WorldHealthOrganisa tion,UNICEF ,Geneva,2008.

2 DepartmentforT rade andIndustry.24th (Final)report ofthehomeand leisure

accidentsurve illancesys tem.2000,2001and2002data. London:Departm entfor

TradeandIndus try, 2003.

3 Healthand SocialCar eInforma tionCentre. HospitalEpisodeStatis tics,Admitted

PatientCare, England2012-13: Externalcauses. http://www .hscic.gov.uk/ catalogue/PUB12566/hosp-epis-stat-admi-ext-caus-2012-13-tab.xlsx(accessed27

Feb2015).

4 ZielinskiAE, Rochette LM,SmithGA.Stair-r elated injuriesto young childrentreated

in USemergency departments,1999 -2008.Pediatrics2012;129:721-7.

5 YoungB,Wynn PM,He Z,et al. Preventingchildhoodfallswithin thehome :

Overviewofs ys tematicreviewsandasys tematicreviewof primarys tudies. Accid

Anal Prev2013;60:158-71.

6 KendrickD, MaulaA, Stew artJ, et al. Keepingchildren safeathome: protocolfor

threema tchedcase-controls tudiesofmodifiable riskfa ctorsforfalls.Inj Prev

2012;18:e3.7 EdwardsPJ,R obertsI, ClarkeMJ,et al. Methodsto increase responsetopos taland

electronicquestion naires.CochraneDatabase SystR ev2009;3:MR000008.

8 ClampM, KendrickD. Ar andomisedcontr olledtrial ofgeneralpr actitionersa fety

advice forfa milieswithchildren under5 years.BMJ1998;316:1576-9.

9 WatsonM,KendrickD, Coupland C,et al. Providingchildsafety equipmentto

preventinjuries:r andomisedcontr olledtrial.BMJ2005;330:178-81.

10 DepartmentforEduca tion.Educa tionandskillsin yourarea. Pos tcode

distances.http:/ /www.education.gov.uk/cgi-bin/inyourarea/distance.pl(accessed

3 Feb2014).

11 DepartmentforC ommunitiesand LocalGovernmen t.English Indicesof Deprivation

2010. https://www.gov.uk/government/statistics/english-indices-of-deprivation-2010

(accessed27 Feb 2015).

12 TextorJ,HardtJ,Knüppel S.DAGitty: agr aphicaltool foranalyzing causal

diagrams.Epidemiology2011;22:745.

13 BjellandI, DahlAA, HaugTT ,et al. Thevalidity ofthe HospitalAnxiety and

DepressionScale:an upda tedliter aturereview .J PsychosomRes 2002;52:69-77.

14 CrnicKA,Gre enbergMT .Minorparenting stresseswith young children.Child Dev

1990;61:1628-37.

15 CrnicKA,Booth CL.Mothers 'and fathers'perceptionsof dailyhassles ofpar enting

acrossearlychild hood.J MarriageF am1991;53:1043-50.

16 PutnamSP ,GartsteinMA, RothbartMK. Measurement offine-grainedasp ectsof

toddlertemper ament:theearlychildhood behavior questionnaire. Infant BehavDev

2006;29:386-401.

17 Putnam SP,Rothbart MK.Developmentof shortand veryshortforms ofthe children's behaviorquestionnair e.J PersAssess2006;87:102-12.

18 GartsteinMA,R othbartMK. Studyinginfanttemperament viathe Revised Infant

Behavior Questionnaire.Infant BehavDev2003;26:64-86.

19 RockhillB,Ne wmanB, WeinbergC.Use andmisuse ofpopulationa ttributable

fractions.Am JPublic Health1998;88:15-19.

20 GLAss essment.TheHospitalAnxiety andDepr ession Scale.Fr equentlyask ed

questions.Ho wshouldmissingda tabe trea ted.http://www.gl-assessment.co.uk/ products/hospital-anxiety-and-depression-scale/hospital-anxiety-and-depression- scale-faqs(a ccessed27Feb 2015).

21 RubinDB. Multiple imputationfornonr esponsein surveys . NewYork: JohnWiley &

Sons, 1987.

22 PublicHealth England. Healthprofiles 2014.http:/ /www.apho.org.uk/default.aspx?

RID=49802(ac cessed13May2015).

23 GielenA C,ShieldsW, McDonaldE, et al. Homesa fetyandlow-income urban

housing quality.Pediatrics2012;130:1053-9.

24 ElkingtonJ,Blogg S,Kelly J,et al. Headinjuries ininfants: acloser looka t

baby-walkers,stairsand nurseryfurniture. N SW PublicHealth Bull1999;10:82-3.

916Kendrick D,et al.ArchDis Child2016;101:909-916. doi:10.1136/archdischild-2015-308486

Original article

on August 11, 2023 by guest. Protected by copyright.http://adc.bmj.com/Arch Dis Child: first published as 10.1136/archdischild-2015-308486 on 1

0 December 2015. Downloaded from