A Classification and Regression Tree for Predicting

A Classification and Regression Tree for Predicting Recurrent Falling among Communitydwelling Seniors Using Home-care Services Author(s): Bernard S. Leclerc, Claude Bégin, Élizabeth Cadieux, Lise Goulet, Jean-François Allaire, Julie Meloche, Nicole Leduc and Marie-Jeanne Kergoat Source: Canadian Journal of Public Health / Revue Canadienne de Sante'e Publique, Vol. 100, No. 4 (July/August 2009), pp. 263-267 Published by: Canadian Public Health Association Stable URL: http://www.jstor.org/stable/41995265 . Accessed: 17/06/2014 21:58 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

.

Canadian Public Health Association is collaborating with JSTOR to digitize, preserve and extend access to Canadian Journal of Public Health / Revue Canadienne de Sante'e Publique.

http://www.jstor.org

This content downloaded from 195.78.108.20 on Tue, 17 Jun 2014 21:58:05 PM All use subject to JSTOR Terms and Conditions

RESEARCH QUANTITATIVE A Classification among

and

Tree

Regression

Community-dwelling

Seniors

for

Predicting

Using

Recurrent

Home-care

Falling

Services

Bernard S. Ledere, MSc,1'*Claude Bégin, MSc,2Élizabeth Cadieux, MSc,1Lise Goulet, MD, PhD,3JeanFrançois Allaire, MSc,4Julie Meloche, MSc,4Nicole Leduc, PhD,3Marie-Jeanne Kergoat, md, CCFP,fcfp,CSPQ5

ABSTRACT Aprospective, observational wasundertaken toidentify risk ofsubjects therecurrence offalling Objectives: study profiles regarding among seniors services. community-dwelling usinghome-care Methods:Aconvenience of868community-dwelling olderpersons, orolder, whousehome-care services offered sample aged65 years bypublic centres intheprovince ofQuébec.Subjects wererecruited between 2002and2005,assessed forfall-related risk and factors, community-based monitored forprospective falls. Datawereexamined andregression tree(CART) andsurvival bya classification analyses. Results:Ninety-nine twofalls within sixmonths ofentry tothestudy. ofrecurrent fallers was11.4%.Thetree Thus,theincidence participants reported classified thepopulation intofivegroups inrisk ofrecurrent basedonhistory offallsinthethree months totheinitial analysis differing falling, prior andusualalcohol inthesixmonths from Therelative risks varied interview, score, Bergbalance typeofhousing, consumption preceding study entry. 0.7to5.1. Thesurvival showed thatthelength oftimebefore a recurrent faller varies analysis becoming amongrisk profiles. Conclusion: Thestudy theconstruction ofeasily risk ofrecurrent Thesecanguideclinicians andpublic health permitted interpretable profiles falling. toidentify individuals andtodecideontheappropriate intervention andfollow-up. practitioners high-risk homecareservices; classification risk risk survival falls; health; factors; assessment; Keywords:Accidental elderly; multiple analysis; prognosis; public analysis Latraduction setrouve durésumé à lafin del'article. 100(4):263-67. Con Health 2009; JPublic

30% ofcommunity-dwelling persons, aged65 or older,fallat leastonce peryear,and about15% sustain Approximately multiplefalls.14Multiplefalls are associatedwith an increasedriskofinstitutionalization and death.45 In additionto fallscan reduceself-confidence, and injury,recurrent mobility, socialcontacts.6 Numerousfactors to falls.7Somecan be cormightcontribute rectedand,thus,theeventcanbe avoided.Themostefficient interventionsarethosewhichtargetscreenedfallerswiththehighest riskoffallingagain,ratherthanelderlypeople identified indiscriminately.4'78 The increasing numberof elderlypeople is leadingto greater demandforhome-care services. fallsamongcommunityPreventing seniors services hasbecomea priority in dwelling usinghome-care 9Nonetheless, riskfactors forfalling areoverlooked in this Québec.7 6 specific population.4 Clinicians areinterested inpredicting adverseoutcomes. Theaim ofthisstudyhasbeento developprofiles forpredicting theriskof recurrent and regression tree-based falling,usinga classification survival analysis.

homenursing services becauseofa care,personalcareand support or a loss of functional temporary disability autonomy.10 Peoplewho could speakneitherFrenchnor English,thosenot able to walk morethansixmetres, andthosewithreducedcommunication and to the Functional Measurement cognitionaccording Autonomy wereexcluded.Allsubjectsgaveinformed consent.The System10 was the authorities of each centre. study approvedby participating Author Affiliations 1. Service desurveillance, recherche etévaluation, Direction desanté et publique delasanté etdesservices sociaux deLanaudière, d'évaluation, joliette, Agence ofthestudy) QC(atthetime 2. Service de prévention et de promotion, Direction de santépublique et delasanté etdesservices sociaux deLanaudière, d'évaluation, joliette, Agence QC 3. Groupe derecherche ensanté, Université deMontréal, Montréal, interdisciplinaire QC 4. TheStatistics Institut deMontréal, QC Montréal, Consulting Group, Philippe-Pinel 5. Research Institut universitaire degériatrie deMontréal, Centre, Montréal, QC * Thisresearch ispart ofLeclerc's PhDthesis inPublic Health andEpidemiology, realized under thesupervision ofProfessors LiseGoulet andNicole Leduc, from theDépartement demédecine sociale etpréventive andthe respectively d'administration delasanté, Faculté demédecine, Université de Département Montréal, Montréal, QC,Canada. Bernard-Simon Direction desindividus Ledere, Correspondence: Développement etdescommunautés, Institut national desanté duQuébec, 190,boul. publique Crémazie Tel:514-864-1600, ext.3530, Fax:514-864est,Montréal, QC H2P1E2, 5190,E-mail: [email protected]. METHODS Theauthors wish tothank allolder clients andhealth care Acknowledgements: workers from thecommunity health andsocial inLanaudière service centres for their Setting and subjects inthestudy. Wealsoacknowledge thecontribution ofGeneviève participation The samplestudiedherewas a conveniencesampleofvolunteers Marquis for thedataentry, for thedataprocessing, andBruce Charles joséePayette Bezeau for therevision ofthemanuscript. Theresearch wassponsored bytheAgence recruited betweenMarch2002 andJuly2005 amongcommunity- de lasanté etdesservices sociaux de Lanaudière andtheGroupe derecherche ensanté oftheUniversité deMontréal. livingpersons,aged65 yearsor older,who werereceiving public interdisciplinaire ©Canadian Public Health 2009.Allrights reserved. Association,

• JULY/AUGUST CANADIAN OFPUBLIC HEALTH 2009263 JOURNAL

This content downloaded from 195.78.108.20 on Tue, 17 Jun 2014 21:58:05 PM All use subject to JSTOR Terms and Conditions

ATREEFORPREDICTING RECURRENT FALLERS elsewhere.11 Ofthe Additional detailsareprovided methodological andagreedtopar959persons whometthestudyinclusioncriteria 868 participants wereusedin theanalyses(Figure1). ticipate, Assessmentof falls and predictors in thesubjectinadvertentA fallwasdefinedas an eventresulting otherlowerlevel.Excludrest on the or to floor, ground, lycoming falls.1The outcomewas measuredby ed were sports-related A fallscalenusingmonthly telephonequestionnaire. self-report each time to individuals to mark events darwaspreviously given had fallen twice fallers were who Recurrent subjects theyappear. withinthefirst sixmonthsoffollow-up.312 ofrecurrent Potential fallingand subjects'characterpredictors at baselineat home.Numberoffallsin the isticswereascertained priorthreemonthswas categorizedas 0, 1, or >2. Nutritional on a graded13-pointscale to identify was performed screening intakedeficienindividuals at highriskofenergyand nutritional as follows: were defined cies.Pre-established 0-2,3-5,and categories 1314 and was 6-13. Bodyweightwasself-reported height measured. BMIvaluesweredefinedas 30. Gaitand balance on a 56-pointscale(45), andbytheTimedUp & Go test1819 tasks overalltime,in seconds,to completea seriesof functional valuesusedin thestudyarethose (30). Thecutoff tool. ofeach clinicalriskassessment proposedbythedevelopers Data abouttheuse ofbenzodiazepines(yes/no)and numberof fromthecontaindrugswererecorded dailyconsumedprescribed was obtainedaccordingto ofalcoholconsumption ers.A history 21 du Québecquestionnaire.20 de la statistique theInstitut Responses werecategorizedforboth drinkingin the precedingweek (yes/no)and usual drinkingduringthe last six months(non