A correspondence analysis revealed frailty deficits aggregate and are multidimensional

A correspondence analysis revealed frailty deficits aggregate and are multidimensional
of 8
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  A correspondence analysis revealed frailty deficits aggregate and aremultidimensional Nadia Sourial a,1 , Christina Wolfson a,b,c,1, *, Howard Bergman a,d,e,1 , Bin Zhu b ,Sathya Karunananthan a , Jacqueline Quail b,c , John Fletcher a , Deborah Weiss b,c ,Karen Bandeen-Roche f  , Franc¸ois B  eland a,d,e a Solidage Research Group, Centre for Clinical Epidemiology and Community Studies, Jewish General Hospital, Montreal, Quebec, Canada b  Division of Clinical Epidemiology, McGill University Health Centre, 1025 Pine Avenue West, Suite P2.028, Montreal, Quebec, Canada c  Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Montreal, Quebec, Canada d  Division of Geriatric Medicine, Jewish General Hospital, McGill University, 3755 Cote-Sainte-Catherine, Montreal, Quebec, Canada e  Department of Health Administration, Universit   e de Montr   eal, Montreal, Quebec, Canada f   Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA Accepted 6 August 2009 AbstractObjective:  To examine the relationships among seven frailty domains: nutrition, physical activity, mobility, strength, energy, cognition,and mood, using data from three studies. Study Design and Setting:  Data from three studies were separately analyzed using multiple correspondence analysis (MCA). Thegraphical output of MCA was used to assess (1) if the presence of deficits in the frailty domains separate from the absence of deficitson the graph, (2) the dimensionality of the domains, (3) the clustering of domains within each dimension, and (4) their relationship withage, sex, and disability. Results were compared across the studies. Results:  In two studies, presence of deficits for all domains separated from absence of deficits. In the third study, there was separation inall domains except cognition. Three main dimensions were retained in each study; however, assigned dimensionality of domains differed.The clustering of mobility with energy and/or strength was consistent across studies. Deficits were associated with older age, female sex,and disability. Conclusion:  Our results suggest that frailty is a multidimensional concept for which the relationships among domains differ accordingto the population characteristics. These domains, with the possible exception of cognition, appear to aggregate together and share a commonunderlying construct.    2010 Elsevier Inc. All rights reserved. Keywords:  Frailty; Domains of frailty; Older persons; Multiple correspondence analysis; Association; Epidemiology 1. Introduction Frailty is generally acknowledged to be a state of decreased reserve and cumulative decline in multiple phys-iologic systems, resulting in an increased risk of adverseoutcomes [1 e 6]. Nevertheless, there remains debate onits characteristics [2]. Many studies have reported on thepredictive validity of various operational definitions of frailty [3,5 e 8]. However, there has been little researchexploring the relationships among the proposed characteris-tics. Bandeen-Roche et al. [9] delineated underlying classesof individuals based on the characteristics proposed byFried et al. [3]. We are unaware, however, of publishedresearch that has explicitly examined the relationshipsamong the proposed characteristics. Evidence of these rela-tionships is necessary to elucidate whether particular char-acteristics belong to the construct of frailty.The objective of this study was to explore the relation-ships among seven frailty domains using multiple corre-spondence analysis (MCA). To assess the consistency of the findings, the methods were replicated using data fromthree studies. The authors report no conflicts of interest. 1 Nadia Sourial, Christina Wolfson, and Howard Bergman contributedequally to the design and conduct of the study and to the preparation of the article.* Corresponding author. Tel.: 514-934-1934 ext. 44739; fax: 514-934-4458. E-mail address : (C. Wolfson).0895-4356/$  e  see front matter    2010 Elsevier Inc. All rights reserved.doi: 10.1016/j.jclinepi.2009.08.007Journal of Clinical Epidemiology 63 (2010) 647 e 654  What is new? Key finding e  O`ur results suggest that the proposed domains of frailty (nutrition, physical activity, mobility,strength, energy, cognition, and mood), with thepossible exception of cognition, appear to aggre-gate together and are multidimensional.What this adds to what was known? e  Much debate still exists over the characteristicsof frailty. Although some studies have reportedon the predictive validity of various frailtydefinitions, there has been little research exploringthe relationship among individual frailtycharacteristics. e  To our knowledge, this is the first study to explic-itly elucidate the relationships among potentialfrailty characteristics.What is the implication, what should change now? e  More research is needed to explore the relation-ships between these seven proposed domains of frailty and the presence of possible subgroups of frailty. e  This exploratory study is the first phase of a larger-study, the International Database Inquiry onFrailty, where the methodology will be replicatedusing data from 11 different studies to investigatethe consistency of the findings in otherpopulations. 2. Methods 2.1. Description of studies Data collected at baseline were taken from the MontrealUnmet Needs Study (MUNS) [10], the Canadian Study of Health and Aging (CSHA) [11], and the System of Inte-grated Services for Older Persons study (French acronym,SIPA) [12]. The MUNS sample consisted of community-dwelling persons aged 75 years and over living in theMontreal area, with no more than mild cognitive impair-ment. The CSHA was a study of dementia conducted on10,263 institutionalized and community-dwelling Cana-dians aged 65 years and older. For the current analysis,only the CSHA community-dwelling participants whocompleted a clinical assessment were retained. The SIPAstudy was conducted on community-dwelling persons,aged 65 years and older, living on the island of Montrealand with disability in at least one instrumental activity of daily living. 2.2. Domains of frailty OnthebasisofaliteraturereviewcarriedoutaspartoftheCanadian Initiative on Frailty and Aging [13], an inclusivelistoffrailtycriteriawasgenerated.Anexpertpanelgroupedthecriteriaintobroaderdomainsusingterminologyproposedby Bergman et al. [2], Ferrucci et al. [14], and Studenski et al. [15]. Seven domains were retained based on clinicaland biological plausibility: nutrition, physical activity,mobility, strength, energy, cognition, and mood. 2.3. Measures of frailty Two authors independently abstracted potential mea-sures for each frailty domain from each of the three studies.Measures of disability as defined by Katz [16] wereexcluded. The final measures were selected based on howclosely they operationalized the domain, their validity,and their clinical relevance (Table 1).There was no measure of nutrition in SIPA and no mea-sure of strength in MUNS. Given the differences in ques-tions selected for the same domain in each study, wefound it necessary to dichotomize the questions to arriveat a common set of questions and to facilitate interpretation.Whenever possible, dichotomization was based on refer-enced cutoffs. Otherwise, cutoffs were defined favoringsensitivity over specificity in identifying subtle vulnerabil-ity. When more than one measure was available for a givendomain, the overall domain was scored positive for a deficitif any measure was positive. 2.4. Analysis Bivariate correlations between the frailty domains wereexamined using the tetrachoric correlation coefficient(TCC). The graphical output of the MCAwas used to assessif the presence of deficits in the frailty domains separatefrom the absence of deficits on the graph; the dimensional-ity of the domains; the clustering of domains within eachdimension; and their relationship with age, sex, and disabil-ity. A detailed description of this method is presented inSourial et al. [17]. Briefly, MCA is used to identify a smallnumber of dimensions in which the largest deviations fromindependence can be represented. These dimensions areidentified through decomposition of the total inertia (afunction of the chi-square statistic). Dimension 1 representsthe greatest deviation from independence in the data; di-mension 2, the second greatest deviation and so on. Eachdimension can be interpreted based on how the presenceand absence of deficit for the domains separate on eitherside of the dimension. Moreover, the further away fromthe srcin a domain’s response category is along a certaindimension, the greater its importance in interpreting that di-mension. This positioning also provides insight into the di-mensionality of the response categories, which group or‘‘load’’ together on a same dimension. ‘‘Clusters’’ orresponse categories very close together on the graph and 648  N. Sourial et al. / Journal of Clinical Epidemiology 63 (2010) 647  e 654  loading on a same dimension tend to indicate similaritiesbetween the responses with moderate to high correlations.Although related, dimensionality of domains differs fromclustering because domains can load on the same dimen-sion without necessarily clustering very closely together.Because domain variables are dichotomous, the two pointsfor presence and absence of deficit in each domain are sym-metric around the srcin. Consequently, the dimensionalityand clustering of points is the same whether looking at thepresence or the absence of deficit. Given this symmetry,throughout the article, we refer to dimensionality and clus-tering of the  domains  rather than the  domain categories .Because we plan to study disability in activities of dailyliving (ADL) as an outcome of frailty in subsequentanalyses, the main analysis was carried out only using thesubsamples without disability. For comparison, however,MCA was repeated using all subjects. ADL disability wasdefined as being unable or needing help to eat, dress, trans-fer, bathe, or toilet [16]. The relationship between thedomain variables and age (65 e 74, 75 e 84, and 85 þ years),sex, and disability was investigated using all the subjects byincluding the variables in the MCA as supplementary vari-ables. The MCA was performed using PROC CORRESP(SAS 9.1, Cary, NC). Multiple imputation (MI) was usedfor missing data [18] and was performed using the Imputa-tion and Variance Estimation Software [19]. 3. Results Table 2 describes the demographic and frailty character-istics of the three samples. There were 839 participants inMUNS, a subsample of 1,600 from CSHA and 1,164 inSIPA.The proportion of missing data in the proposed frailtydomains varied across the three samples. No data weremissing in the MUNS. In SIPA, domains had less than2% missing data with the exceptions of cognition (8.3%)and mood (17.9%). In the CSHA, less than 4% of the datawere missing except for physical activity (13.8%). Age,presence of disability, and the average number of frailtydeficits were lowest in MUNS and highest in SIPA. There Table 1Measures of the frailty domainsFrailty domain MUNS CSHA SIPANutrition Self-reported:1. Weight loss in thepast year, or2. Medium or poor appetiteSelf-reported a :1. Loss of appetite, or2. Unintentional weight loss, or3. Unexplained weightchange O 3 kg over 6 moNAPhysical activity Self-reported never or fromtime to time:1. Participating in a leisuretime physical activity group, and2. Doing a hobby (sewing, gardening,reading, do-it-yourself projects)in the past monthSelf-reported low or no exercise Self-reported < 4 times/monthgoing to a:1. Shopping center or other placesto walk around, and2. Recreation or cultural centerMobility Self-reported:1. Difficulty or needing helpmoving around the house, or2. Therapy to improve balanceand mobilityQuestionably or definitivelyabnormal gait speed a Self-reported difficulty walkingup or down a flight of stairsEnergy Self-reported difficulty, needinghelp or inability to walk a block Participant feeling tiredall the time a Self-reported either:1. Difficulty walking one mile, or2. Needing help or unable to walk a block during summerStrength NA Questionably or definitivelyabnormal a :1. Strength, or2. Bulk Self-reported difficulty liftingsomething that weighs over 5 kgCognition  < 17 on the ALFI [26] Diagnosis of cognitiveimpairment or dementia accordingto DSM-III-R criteria a 1. Self-reported memory problemsor Alzheimer’s disease, or2.  > 3 errors on the SPMSQMood  O Median (11.9) on theIDPESQ-141. Questionable or definitivedepression, a or2. Participant feeling sad, blue,or depressed a 1. Self-reported emotional problems,nervousness, depression, anxiety,or insomnia, or2.  > 5 on the GDS  Abbreviations : MUNS, Montreal Unmet Needs Study; CSHA, Canadian Study of Health and Aging; SIPA, Integrated Services for Older Persons; NA, notavailable; ALFI, Adult Lifestyles and Function Interview; IDPESQ-14,  Indice de d   etresse psychologique de Sant   e Qu  ebec ; DSM-III-R, Diagnostic StatisticalManual of Mental Disorders; SPMSQ, Short Portable Mental Status Questionnaire; GDS, Geriatric Depression Scale. a Determined during physician assessment as part of a medical history.649  N. Sourial et al. / Journal of Clinical Epidemiology 63 (2010) 647  e 654  were more women than men in all the three samples. Thepercentage of subjects with deficits in mobility, energy,strength, and mood was highest in SIPA, whereas the per-centage of subjects with poor physical activity and cogni-tive deficits was highest in CSHA. The most commondeficits in MUNS were nutrition and mood; in CSHA,physical activity and cognition; and, in SIPA, physicalactivity and energy.Three main dimensions were identified in each study.The proportion of inertia explained by dimensions 1, 2,and 3 was 64.8% in MUNS, 65.4% in CSHA, and 75.5%in SIPA. Fig. 1 presents the results among the disability-free participants. In MUNS and the CSHA, there was com-plete separation between presence and absence of deficitson either side of dimension 1. This suggests that the mostimportant difference in the samples was between havingfrailty markers and not, indicating a positive correlationbetween these domains. However, cognition was onlyweakly associated with the other domains in MUNS, asseen by the relatively wide angle between cognition andthe others in Fig. 1. This weak association was also sup-ported by the low TCCs ranging from   0.05 to 0.14 forthe correlation between cognition and the other domains(Table 3). In SIPA, cognition did not separate to the sameside as the other domains. Indeed, its correlation with theother domains was found to be negligible and in some casesnegative (TCCs 5  0.25 to 0.00).Table 4 presents the dimensionality and clustering of domains across the three studies. Frailty domains enumer-ated on the same line for a given dimension indicate a mod-erate to high correlation and clustering of these domains onthe graph. In MUNS, separation of presence from absenceof deficits in energy and mobility appears to be the mostimportant in explaining the deviation from independencein this sample given their strong representation ondimension 1. Moreover, these two domains form a closecluster on the graph (TCC 5 0.57). Cognition and, toa lesser degree, nutrition load best onto dimension 2 butare almost 90 degrees apart from the srcin indicatinga weak relationship (TCC 5  0.05). Physical activityand, to a lesser extent, mood are best represented on dimen-sion 3 and are moderately correlated (TCC 5 0.22). InCSHA, dimension 1 is mainly characterized by mood, nu-trition, and energy, which cluster closely together in the up-per right quadrant. (TCC 5 0.44 e 0.62). Cognition does notcontribute much information given its relatively low load-ing on all three dimensions, although slightly higher on di-mension 1. Along dimension 2, mobility and strength havethe most importance and also form a cluster (TCC 5 0.59).Physical activity seems to form a separate dimension, load-ing on dimension 3. In SIPA, energy, mobility, and strengthare the most important on dimension 1 and cluster veryclosely on the graph given their strong correlation (TCC 5 0.48 e 0.72). Physical activity, although close to energy,mobility, and strength on dimension 1, had a slightly highercontribution on dimension 3. Cognition and mood werewell represented on dimension 2 and dimension 3 but werestrongest on dimension 2. These two domains, however,did not cluster on the graph and were not correlated(TCC 5 0.00).Findings using all the subjects in each study (i.e., includ-ing thosewith ADL disability) were similar, with minor var-iations in coordinates of responses on the graph. In SIPA,where the proportion with ADL disability was highest, usingthe full sample sometimes resulted in a different assignmentof dimensionality for domains which contributed to morethan onedimension,for example,inSIPAusingthefullsam-ple, physical activity loaded on both dimensions 1 and 3 butwasslightlyhigherondimension1(resultsnotshown).Whenadding age, sex, and disability to the analysis using all thesubjects, the presence of frailty deficits was found, in allthe three studies, to be associated with older age, femalesex, and the presence of disability. The most important sepa-ration in relation to age was between the participants agedyounger than 85 years and those aged 85 years and older,the latter being associated with the presence of deficits. 4. Discussion To the best of our knowledge, this is the first study toexplicitly investigate the relationships among frailtydomains. The replication of the methodology in samplesfrom three elderly target populations allowed us to assessthe consistency of the findings.If the proposed frailty domains aggregate as a syndrome,a necessary although not sufficient condition is that the pres-ence of deficits in all the domains aggregate together on thesame side of dimension 1 on the correspondence analysisgraph,separatedfromtheabsenceofdeficitsontheothersideof dimension 1. This complete separation of presence from Table 2Sample characteristicsMUNS( n 5 839)CSHA( n 5 1600)SIPA( n 5 1164)AgeMean (SD) 79.6 (3.9) 80.1 (6.9) 82.2 (7.3)Range 74 e 96 65 e 99 64 e 104Female (%) 68.7 58.9 70.9Any ADL disability (%) 5.5 26.1 42.4Mean number of frailty deficitsper subject (SD)1.7 (1.2) 2.2 (1.6) 4.5 (1.3)Percentage (%) with deficitNutrition 52.3 16.8 NAPhysical activity 20.5 67.7 57.5Mobility 16.0 33.6 75.0Energy 19.4 10.4 87.3Strength NA 16.9 74.2Cognition 6.2 53.3 52.8Mood 51.3 22.0 71.2  Abbreviations : MUNS, Montreal Unmet Needs Study; CSHA, Cana-dian Study of Health and Aging; SIPA, Integrated Services for Older Per-sons; SD, standard deviation; NA, not available.650  N. Sourial et al. / Journal of Clinical Epidemiology 63 (2010) 647  e 654  absence ofdeficits across alldomainsoccurredintheMUNSand CSHA studies, although the aggregation of cognitionwith the other domains was weak in MUNS. In SIPA, therewas separation in all the domains except cognition. In allthethreestudies,MCAproducedverysimilarresultswhetheror not we included participants with ADL disability, MUNS PA_0PA_1MB_0MB_1NU_0NU_1EN_0EN_1CG_0CG_1MD_0MD_1 -1,5-1-0,500,511,522,533,54 Dimension 1 (31.6%)    D   i  m  e  n  s   i  o  n   3   (   1   6 .   2   %   ) MUNS PA_0PA_1MB_0MB_1NU_0NU_1EN_0EN_1CG_0CG_1MD_0MD_1 -1,5-1-0,500,511,522,533,54 Dimension 1 (31.6%)    D   i  m  e  n  s   i  o  n   2   (   1   7 .   0   %   ) CSHA PA_0PA_1MB_0MB_1ST_0ST_1NU_0NU_1EN_0EN_1CG_0CG_1MD_0MD_1 -1,5-1-0,500,511,522,533,54 Dimension 1 (33.9%)    D   i  m  e  n  s   i  o  n   2   (   1   8 .   7   %   ) CSHA PA_0PA_1MB_0MB_1ST_0ST_1NU_0NU_1EN_0 EN_1CG_0CG_1MD_0MD_1 -1,5-1-0,500,511,522,533,54 Dimension 1 (33.9%)    D  v   i  m  e  n  s   i  o  n   3   (   1   2 .   8   %   ) SIPA PA_0PA_1MB_0MB_1ST_0ST_1EN_0EN_1CG_0CG_1MD_0MD_1 -1,5-1-0,500,511,522,533,54 Dimension 1 (44.1%)    D   i  m  e  n  s   i  o  n   2   (   1   6 .   5   %   ) SIPA PA_0PA_1MB_0 MB_1ST_0ST_1EN_0EN_1CG_0CG_1MD_0MD_1 -2-1,5-1-0,5 0 0,5 1 1,52-1 -0,5 0 0,5 1 1,5 2-1 -0,5 0 0,5 1 1,5 2-1 -0,5 0 0,5 1 1,5 2 2,5 -1 -0,5 0 0,5 1 1,5 2 2,5-2 -1,5 -1 -0,5 0 0,5 1 1,5 2 -2 -1,5 -1 -0,5 0 0,5 1 1,5 2 Dimension 1 (44.1%)    D   i  m  e  n  s   i  o  n   3   (   1   4 .   9   %   ) Fig. 1. Relationship among the frailty domain variables across the three studies d results of the multiple correspondence analysis. MUNS, Montreal UnmetNeeds Study; CSHA, Canadian Study of Health and Aging; SIPA, Integrated Services for Older Persons; NU, nutrition; PA, physical activity; MB, mobility;ST, strength; EN, energy; CG, cognition; MD, mood; suffix 1 5 presence of deficit; suffix 0 5 absence of deficit; percentages for each axis correspond toproportion of explained inertia in each dimension.651  N. Sourial et al. / Journal of Clinical Epidemiology 63 (2010) 647  e 654
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks