Age and Ageing 2021; 50: 2183–2191 © The Author(s) 2021. Published by Oxford University Press on behalf of the British Geriatrics https://doi.org/10.1093/ageing/afab138 Society. All rights reserved. For permissions, please email: journals.permissions@oup.com Published electronically 3 July 2021 This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com RESEARCH PAPER Multimorbidity patterns and risk of frailty in older community-dwelling adults: a population-based cohort study Clare Tazzeo1 , Debora Rizzuto1,2 , Amaia Calderón-Larrañaga1 , Albert Roso-Llorach3,4 , Alessandra Marengoni1,5 , Anna-Karin Welmer1 , Graziano Onder6 , Caterina Trevisan7,8 , Davide Liborio Vetrano1,9 1 Aging Research Center, Department of Neurobiology, Care Sciences and Society, Karolinska Institutet and Stockholm University, Stockholm, Sweden 2 Stockholm Gerontology Research Center, Stockholm, Sweden 3 Fundació Institut Universitari per a la recerca a l’Atenció Primària de Salut Jordi Gol i Gurina (IDIAPJGol), Barcelona, Spain 4 Universitat Autònoma de Barcelona, Campus de la UAB, Bellaterra (Cerdanyola del Vallès), Barcelona, Spain 5 Department of Clinical and Experimental Sciences, University of Brescia, Brescia, Italy 6 Department of Cardiovascular, Endocrine-Metabolic Diseases and Aging, Istituto Superiore di Sanità, Rome, Italy 7 National Research Council-Neuroscience Institute, Aging Branch, Padova, Italy 8 Geriatric Unit, Department of Medicine (DIMED), University of Padova, Padova, Italy 9 Centro Medicina dell’Invecchiamento, Fondazione Policlinico Universitario "A. Gemelli" IRCCS, and Università Cattolica del Sacro Cuore, Rome, Italy Address correspondence to: Clare Tazzeo, Aging Research Center, Karolinska Institutet, Tomtebodavägen 18 A, 9th Floor, Solna 171 65, Sweden. Tel.: +46 8 524 858 39. Email: clare.tazzeo@ki.se Abstract Background: the aim of this study was to examine the cross-sectional and longitudinal associations of different multimorbidity patterns with physical frailty in older adults. Methods: we used data from the Swedish National study on Aging and Care in Kungsholmen to generate a physical frailty measure, and clusters of participants with similar multimorbidity patterns were identified through fuzzy c-means cluster analyses. The cross-sectional association (n = 2,534) between multimorbidity clusters and physical frailty was measured through logistic regression analyses. Six- (n = 2,122) and 12-year (n = 2,140) longitudinal associations were determined through multinomial logistic regression analyses. Results: six multimorbidity patterns were identified at baseline: psychiatric diseases; cardiovascular diseases, anaemia and dementia; sensory impairments and cancer; metabolic and sleep disorders; musculoskeletal, respiratory and gastrointestinal diseases; and an unspecific pattern lacking any overrepresented diseases. Cross-sectionally, each pattern was associated with physical frailty compared with the unspecific pattern. Over 6 years, the psychiatric diseases (relative risk ratio [RRR]: 3.04; 95% confidence intervals [CI]: 1.59–5.79); cardiovascular diseases, anaemia and dementia (RRR 2.25; 95% CI: 1.13–4.49) and metabolic and sleep disorders (RRR 1.99; 95% CI: 1.25–3.16) patterns were associated with incident physical frailty. The cardiovascular diseases, anaemia and dementia (RRR: 4.81; 95% CI: 1.59–14.60); psychiatric diseases (RRR 2.62; 95% CI: 1.45–4.72) and sensory impairments and cancer (RRR 1.87; 95% CI: 1.05–3.35) patterns were more associated with physical frailty, compared with the unspecific pattern, over 12 years. Conclusions: we found that older adults with multimorbidity characterised by cardiovascular and neuropsychiatric disease patterns are most susceptible to developing physical frailty. Keywords: multimorbidity, frailty, older people, longitudinal population-based study, personalised medicine 2183 C. Tazzeo et al. Key Points • Older multimorbid adults characterised by cardiovascular and neuropsychiatric disease patterns are at higher frailty risk. • Multimorbidity patterns are differentially associated with incident physical frailty. • Not all older adults with multimorbidity are physically frail. Introduction population-based cohort study, following older adults (60+ years), living in Kungsholmen; a district of Stockholm, Multimorbidity is generally understood to be the presence Sweden [27]. Participants were randomly selected from of two or more chronic conditions in one individual [1]. Its 11 age cohorts (60, 66, 72, 78, 81, 84, 87, 90, 93, prevalence in older adults varies from 55 to 98%, depend- 96 and 99+ years [27]). At baseline (2001–2004), 73% ing on the definition utilised and population studied [2]. (n = 3,363) of all eligible, invited persons enrolled [28]. Multimorbidity is associated with lowered quality of life [2], Since baseline, participants have been followed every 6 years disability [2] and mortality [3], as well as extensive healthcare if younger than 78 and every 3 years if 78 or older utilisation [2], prolonged hospital stays [4] and complex [28]. This study was approved by the Ethics Committee pharmacological regimes [5, 6]. at Karolinska Institutet and the Regional Ethics Review Along with multimorbidity, frailty limits health and sur- Board in Stockholm. Participants, or proxy decision-makers vival in old age. Frailty is characterised by an accelerated for those with cognitive impairment, provided informed ageing of organs and bodily systems [7]. When confronted consent. with minor stressors, frail persons experience disproportion- The cross-sectional analyses included 2,534 participants ate negative health outcomes [7]. Frail individuals are at after excluding 190 institutionalised persons, 432 with less increased risk for mortality [8, 9], hospitalisation [7], falls than two diseases and 207 missing frailty information. Those [10], long-term care [8], loneliness [11] and reduced quality missing baseline frailty status were older, more likely to be of life [12]. A 2020 meta-analysis found physical frailty’s female and, on average, affected by a higher number of prevalence to be 12% in adults aged 50 and above [13]. diseases (Supplementary Table S1, Supplementary data are An association between multimorbidity and frailty has available in Age and Ageing online). The same participants been previously established [14–20]. A 2019 meta-analysis were included in the longitudinal analyses, after excluding found that 72% of frail individuals exhibit multimorbidity, those with baseline frailty. In total, 2,122 participants were but only 16% of multimorbid persons have frailty [21]. It included in the 6-year analysis (94 missing frailty status) is unclear why some with multimorbidity maintain robust and 2,140 in the 12-year analysis (76 missing frailty status; health states while others experience frailty. An explanation Supplementary Figure S1, Supplementary data are available could be that most studies have defined multimorbidity by in Age and Ageing online). disease count, which is a poorly discriminative measure [2]. Psychosocial, biological, pharmacological and other health factors heighten one’s likelihood of developing certain dis- Data collection eases and can result in systematic disease clustering [22, The following data were collected at each visit: (i) a physi- 23]. Defining multimorbidity through clusters accounts for cal functioning assessment and social interview, conducted disease patterns and diversity, which are differently associated by a trained nurse; (ii) a clinical examination involving a with negative outcomes such as functional decline [24] and physician conducting neurological, geriatric and psychiatric mortality [25]. A cross-sectional study used cluster analysis assessments and (iii) a psychologist-administered cognitive to examine chronic diseases that exist in robust and frail pop- assessment [28]. ulations of multimorbid individuals [26]. However, there have not been investigations into the association between multimorbidity patterns and incident physical frailty. This Chronic disease assessment study aims to determine the cross-sectional and longitudinal associations between different multimorbidity patterns and Clinical diagnoses in SNAC-K were ascertained by physi- physical frailty in older Swedish adults. cians through clinical examinations, self-reported health, review of medical journals, clinical lab parameters, anamnes- tic data and medication-use. All four-digit level International Methods Classification of Diseases, 10th Revision (ICD-10) codes were designated as chronic or non-chronic, and a list of 60 Study design and population categories of chronic diseases was identified using baseline Data were used from The Swedish National study on SNAC-K data, considering disease prognosis, prevalence, Aging and Care in Kungsholmen (SNAC-K): an ongoing pathophysiology and treatment [29]. 2184 Multimorbidity patterns and risk of frailty in older community-dwelling adults Physical frailty assessment outcome. Participants were then assigned to the cluster in Frailty was operationalised in accordance with a modified which they had the highest membership probability, making version of the frailty phenotype developed by Fried et al., it possible that participants in different clusters could share which identifies an individual as physically frail if they common diseases. To characterize the clusters of individuals exhibit at least three of the following: unintentional weight in terms of diseases, observed/expected ratios were used, loss, low energy expenditure, self-reported exhaustion, slow comparing disease prevalence in the cluster with that of gait speed and weak grip strength [30]. Unintentional weight the sample population. Additionally, disease exclusivity was loss was defined as loss of at least 1 kg within the last calculated as the fraction of individuals with the disease 3 months. Those exercising three times per month or less in a cluster divided by the total number of individuals were said to have low energy expenditure. Self-reported with the disease. Diseases were recognised to be associated exhaustion was defined as reported fatigue within the last with specific clusters when the exclusivity was ≥25% 3 months. Gait speed (meters/second) was timed as par- or the observed/expected ratio was ≥2 (Supplementary ticipants walked 6 meters (m), or 2.4 m for those who Table S2, Supplementary data are available in Age and considered themselves slow walkers. Slow gait speed was Ageing online). Diseases with both an exclusivity ≥25% defined as the slowest 20%, adjusted by height and sex. and observed/expected ratio ≥ 2 were represented in the Participants missing gait speed were classified as slow walkers cluster’s name. Further details on this methodology have if they: could not stand up or walk without assistance, used been published [25, 31]. a wheelchair, could not move around without assistance, or Participants’ baseline characteristics were analysed by could not walk 100–200 m without great difficulty. Those multimorbidity pattern through chi-square tests and one- able to walk 1 km without difficulty were said to have non- way analysis of variance. Logistic regression was utilised slow gait speed. Grip strength (Newtons) was measured in to assess the cross-sectional association between the multi- both hands with an electronic dynamometer (Grippit ), morbidity patterns and baseline frailty. Multinomial logistic using the strongest value of the two. Weak grip strength was regression was used to assess the longitudinal (6- and 12- classified as the lowest 20% of participants, adjusted by sex year) association between the multimorbidity patterns and and body mass index. Those missing a grip strength measure frailty, and the outcome was categorised into non-frail were said to have weak grip strength if they could not open (reference group), frail, dropouts and deceased (using data jars with lids. Individuals missing some of the five frailty from the Swedish Cause of Death Register). All models were criteria were not excluded if their available variables could first adjusted for sociodemographic characteristics, and then classify them as frail (at least three frail criteria) or robust (at for additional demographic (e.g. civil status) and behavioural least three non-frail criteria). factors (e.g. smoking history and alcohol consumption). Stratified analyses by age and sex were performed to test for a potential modifier effect. A P-value <0.05 was considered Covariates statistically significant in each analysis. Stata/IC 15.1 and R Education was categorised into elementary (<8 years), high 4.0.0 were employed for the analyses. school (8–12 years) and university (≥13 years). Civil status was categorised into partnered, widowed, unmarried and Sensitivity analyses divorced. Smoking history was categorised as: never, former Missing data could have resulted in decreased statistical or current. Alcohol consumption was categorised as: never or power and biased estimations of the association between occasional, light to moderate (1–14 standard drinks/week in the multimorbidity patterns and physical frailty. As such, men, 1–7 in women) or heavy (>14 standard drinks/week multiple imputation using chained equations (MICE) was in men, >7 in women). conducted as a sensitivity analysis [32]. Covariates and phys- ical frailty status with missing observations were imputed, Statistical analyses under the assumption that these observations were missing at random. Five imputed datasets were generated, using To identify multimorbidity patterns, all non-institutionalised demographic data and disease status as auxiliary variables. participants with at least two chronic diseases were included. Under Rubin’s rule, estimates were generated by pooling the Upon performing a dimensionality reduction (i.e. multiple results of the imputed datasets [33]. correspondence analysis) accounting for all diseases display- ing a prevalence >2%, a fuzzy c-means cluster algorithm was employed to generate clusters of individuals (based Results on their diseases). This allowed individuals to belong to more than one cluster, as all participants were assigned The mean age of the 2,534 participants in the cross-sectional a membership probability to each cluster. To obtain the analyses was 74.3 ± 10.3 years and 64% were female. optimal cluster number, validation indices and various Six multimorbidity patterns were identified at baseline: degrees of fuzzification were tested. The soft clustering psychiatric diseases (n = 149; 5.9%); cardiovascular diseases, analysis was repeated 100 times to account for the random anaemia and dementia (n = 195; 7.7%); sensory impairments nature of cluster solutions and generate an average final and cancer (n = 287; 11.3%); metabolic and sleep disorders 2185 C. Tazzeo et al. Table 1. Baseline characteristics of complete cases by multimorbidity pattern Characteristics Unspecific MSK, resp. & Metabolic & Sensory & Cardio., anaemia & Psych. Total (n = 2,534) (n = 1,210, GI (n = 402, sleep (n = 291, cancer (n = 287, dementia (n = 195, (n = 149, 47.8%) 15.9%) 11.5%) 11.3%) 7.7%) 5.9%) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Age 71.2 ± 9.2 74.8 ± 10.2 73.0 ± 8.8 83.2 ± 8.5 83.4 ± 8.6 71.2 ± 9.6 74.3 ± 10.3∗∗ Sex (female) 760 (62.8) 310 (77.1) 135 (46.4) 189 (65.9) 128 (65.6) 109 (73.2) 1,631 (64.4)∗∗ Education Elementary 176 (14.6) 66 (16.4) 49 (16.8) 69 (24.0) 56 (28.9) 21 (14.1) 437 (17.3)∗∗ High school 589 (48.7) 202 (50.3) 155 (53.3) 151 (52.6) 102 (52.6) 71 (47.7) 1,270 (50.2) University 444 (36.7) 134 (33.3) 87 (29.9) 67 (23.3) 36 (18.6) 57 (38.3) 825 (32.6)∗∗ Civil status Partnered 637 (52.7) 167 (41.7) 147 (50.5) 84 (29.3) 63 (32.3) 60 (40.3) 1,158 (45.7)∗∗ Widowed 242 (20.0) 107 (26.7) 66 (22.7) 127 (44.3) 86 (44.1) 36 (24.2) 664 (26.2)∗∗ Unmarried 172 (14.2) 63 (15.7) 40 (13.8) 44 (15.3) 30 (15.4) 22 (14.8) 371 (14.7) Divorced 158 (13.1) 64 (16.0) 38 (13.1) 32 (11.2) 16 (8.2) 31 (20.8) 339 (13.4)∗ Alcohol consumption Never/occasional 341 (28.3) 160 (40.1) 100 (34.5) 151 (53.4) 108 (56.5) 53 (35.6) 913 (36.3)∗∗ Light/moderate 651 (54.0) 170 (42.6) 147 (50.7) 97 (34.3) 66 (34.6) 55 (36.9) 1,186 (47.1)∗∗ Heavy 214 (17.7) 69 (17.3) 43 (14.8) 35 (12.4) 17 (8.9) 41 (27.5) 419 (16.6)∗∗ Smoking status Never 559 (46.5) 197 (49.4) 114 (39.3) 160 (56.1) 99 (50.8) 58 (39.5) 1,187 (47.1)∗ Former 457 (38.0) 152 (38.1) 131 (45.2) 94 (33.0) 79 (40.5) 55 (37.4) 968 (38.4) Current 187 (15.5) 50 (12.5) 45 (15.5) 31 (10.9) 17 (8.7) 34 (23.1) 364 (14.5)∗ Number of diseases 3.0 ± 1.1 4.5 ± 1.8 5.1 ± 1.8 5.7 ± 1.9 7.6 ± 2.4 5.4 ± 2.0 4.3 ± 2.1∗∗ Physically frail 56 (4.6) 66 (16.4) 25 (8.6) 69 (24.0) 76 (39.0) 26 (17.5) 318 (12.6)∗∗ Notes: MSK, resp. & GI = musculoskeletal, respiratory & gastrointestinal diseases; metabolic & sleep = metabolic & sleep disorders; sensory & cancer = sensory impairments & cancer; cardio., anaemia & dementia = cardiovascular diseases, anaemia & dementia; psych. = psychiatric. Missing variables: civil status (n = 2), education (n = 2), smoking status (n = 15) and alcohol consumption (n = 16). Values are presented as absolute number and column percentage (%) or mean ± standard deviation. ∗ P < 0.01. ∗∗ P < 0.001 (n = 291; 11.5%); musculoskeletal (MSK), respiratory and Table 3 displays the longitudinal association between the gastrointestinal (GI) diseases (n = 402; 15.9%); and an multimorbidity patterns and incident frailty. Of the 2,122 unspecific (n = 1,210; 47.8%) pattern lacking overrepre- participants included in the 6-year analyses, 236 (11.1%) sented diseases (Supplementary Tables S2 and S3 for details developed frailty, with those belonging to the psychiatric dis- supplementary data are available in Age and Ageing online). eases (relative risk ratio [RRR]: 3.04; 95% CI: 1.59–5.79), Table 1 displays the baseline characteristics of the total popu- cardiovascular diseases, anaemia and dementia (RRR: 2.25, lation (n = 2,534) according to the multimorbidity patterns. 95% CI: 1.13–4.49), and metabolic and sleep disorders Participants in the cardiovascular diseases, anaemia and (RRR: 1.99; 95% CI 1.25–3.16) patterns having statistically dementia pattern (83.4 ± 8.6 years), followed by the sensory significant increased relative risks of developing frailty within impairments and cancer (83.2 ± 8.5 years), were the oldest. 6 years compared with those in the unspecific pattern, in Those in the cardiovascular diseases, anaemia and dementia the fully-adjusted model. Of the 2,140 participants included pattern presented with the highest number of chronic in the 12-year analyses, 435 (20.3%) developed frailty. In diseases (7.6 ± 2.4). The unspecific pattern was relatively the fully adjusted model, those in the cardiovascular dis- healthier, characterised by a younger average age (71.2 ± 9.2) eases, anaemia and dementia (RRR: 4.81; 95% CI: 1.59– and the lowest number of chronic diseases (3.0 ± 1.1). 14.60), psychiatric diseases (RRR: 2.62, 95% CI: 1.45– Overall, 318 (13%) participants with physical frailty 4.72) and sensory impairments and cancer (RRR: 1.87; 95% were identified at baseline, with the highest prevalence in CI 1.05–3.35) patterns had statistically significant increased the cardiovascular diseases, anaemia and dementia pattern relative risks of developing frailty within 12 years compared (39%) and the lowest in the unspecific pattern (5%); with those in the unspecific pattern. See Supplementary prevalence was higher in older age cohorts across all patterns Table S4 for the fully expanded and adjusted longitudinal (Figure 1). In the fully adjusted cross-sectional analyses, all analyses. patterns presented statistically significant associations with There were 2,741, and 2,423, participants included in frailty compared to the unspecific pattern, with odds ratios the cross-sectional and longitudinal analyses, respectively, ranging from 1.74 (95% CI: 1.03–2.91) for the metabolic after employing MICE. The results of the imputed analyses and sleep disorders pattern to 5.23 (95% CI: 3.38–8.09) for align with the complete-case analyses (Supplementary Table the cardiovascular diseases, anaemia and dementia pattern S5, Supplementary data are available in Age and Ageing (Table 2). online). 2186 Multimorbidity patterns and risk of frailty in older community-dwelling adults Figure 1. Baseline physical frailty prevalence and 95% confidence intervals by multimorbidity pattern. Notes: psychi- atric = psychiatric diseases; cardio, anaemia & dementia = cardiovascular diseases, anaemia & dementia; sensory & cancer = sensory impairments & cancer; metabolic & sleep = metabolic & sleep disorders; MSK, resp. & GI = musculoskeletal, respiratory & gastrointestinal diseases. Table 2. Cross-sectional association between multimorbidity patterns and physical frailty in complete cases at baseline (n = 2,534) Pattern Cases Model 1 Model 2 n/N % OR 95% CI OR 95% CI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Unspecific 56/1,210 4.6 1 (ref ) – 1 (ref ) – MSK, resp. & GI 66/402 16.4 3.13 2.10–4.66 3.13 2.08–4.70 Metabolic & sleep 25/291 8.6 1.81 1.09–3.01 1.74 1.03–2.91 Sensory & cancer 69/287 24.0 2.59 1.71–3.93 2.39 1.56–3.66 Cardio., anaemia & dementia 76/195 39.0 5.57 3.63–8.53 5.23 3.38–8.09 Psychiatric 26/149 17.5 4.96 2.90–8.47 4.86 2.80–8.43 Notes: MSK, resp. & GI = musculoskeletal, respiratory & gastrointestinal diseases; metabolic & sleep = metabolic & sleep disorders; sensory & cancer = sensory impairments & cancer; cardio., anaemia & dementia = cardiovascular diseases, anaemia & dementia; OR = odds ratio. Missing observations: civil status (n = 2), education (n = 2), smoking status (n = 15) and alcohol consumption (n = 16). Model 1: adjusted for age, sex and education. Model 2: adjusted for age, sex, education, civil status, smoking status and alcohol consumption. Stratified analyses were conducted and the presence of an Discussion interaction between the patterns and incident physical frailty with age and sex was tested. No interaction was found; based Our results indicate that distinct multimorbidity patterns on the stratified analyses, the association tended to be higher are differentially associated with incident physical frailty in in females, and when stratified by age, significance was lost older adults. Compared with the unspecific one, all patterns except for in the psychiatric diseases pattern, which is likely exhibited a statistically significant cross-sectional association due to power (Supplementary Table S6, Supplementary data with frailty. The psychiatric diseases; cardiovascular diseases, are available in Age and Ageing online). anaemia and dementia; and metabolic and sleep disorders 2187 C. Tazzeo et al. Table 3. Longitudinal association between multimorbidity patterns and physical frailty in complete cases at six (n = 2,122) and 12 (n = 2,140) years Pattern Cases Model 1 Model 2 n/N % RRR 95% CI RRR 95% CI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-year Unspecific 90/1,121 8.0 1 (ref ) – 1 (ref ) – MSK, resp. & GI 38/321 11.8 1.28 0.82–1.99 1.22 0.78–1.92 Metabolic & sleep 35/260 13.5 1.99 1.26–3.15 1.99 1.25–3.16 Sensory & cancer 37/193 19.2 1.31 0.80–2.15 1.36 0.82–2.25 Cardio., anaemia & dementia 20/112 17.9 2.34 1.18–4.62 2.25 1.13–4.49 Psychiatric 16/115 13.9 3.03 1.60–5.74 3.04 1.59–5.79 12-year Unspecific 190/1,119 17.0 1 (ref ) – 1 (ref ) – MSK, resp. & GI 75/323 23.2 1.46 1.00–2.13 1.38 0.94–2.03 Metabolic & sleep 55/260 21.2 1.49 0.97–2.28 1.43 0.93–2.19 Sensory & cancer 61/204 29.9 1.99 1.12–3.52 1.87 1.05–3.35 Cardio., anaemia & dementia 27/116 23.3 5.31 1.75–16.13 4.81 1.59–14.60 Psychiatric 27/118 22.9 2.74 1.54–4.89 2.62 1.45–4.72 Notes: MSK, resp. & GI = musculoskeletal, respiratory & gastrointestinal diseases; metabolic & sleep = metabolic & sleep disorders; sensory & cancer = sensory impairments & cancer; cardio., anaemia & dementia = cardiovascular diseases, anaemia & dementia. Missing observations in 6-year analysis: education (n = 1), civil status (n = 2), alcohol consumption (n = 8) and smoking status (n = 11). Missing observations in 12-year analysis: education (n = 1), civil status (n = 1), alcohol consumption (n = 9) and smoking status (n = 10). Model 1: adjusted for age, sex and education. Model 2: adjusted for age, sex, education, civil status, smoking status and alcohol consumption. patterns were significantly associated with frailty develop- Longitudinally, the results indicate that multimorbidity ment over 6 years compared with the unspecific pattern. patterns characterised by cardiovascular and neuropsychi- Over 12 years, the cardiovascular diseases, anaemia and atric diseases are most strongly associated with physical dementia; psychiatric diseases; and sensory impairments and frailty. The cardiovascular diseases, anaemia and dementia cancer patterns were significantly associated with incident pattern is characterised by cardiac diseases that impact organ frailty compared with the unspecific pattern. perfusion, oxygenation, and thus, muscle performance, The relation between multimorbidity and frailty has been which is strongly implicated in frailty’s pathogenesis [42, reported in several studies [14–21], but others do not sup- 43]. Furthermore, dementia may impact one’s nutritional port this association [34, 35]. Varying definitions of mul- status, physical activity engagement and muscle fitness, timorbidity might be a reason for these conflicting results. promoting frailty development [39, 44]. The neuropsychi- Despite its widespread recent use, and in being in line atric conditions in the psychiatric diseases pattern, such with the deficit accumulation frailty model, operationalizing as depression and anxiety, are associated with fatigue as multimorbidity through disease counts neglects the nature well as reduced appetite and physical activity, which are and interactions of diseases, which may differently affect key components of frailty [30, 40]. Nguyen et al. found frailty susceptibility. We assessed multimorbidity through that neuropsychiatric multimorbidity was associated with patterns, recognizing its qualitative complexity and the ten- increased mortality in robust, but even more so in frail, dency of certain diseases to coexist [22]. Although frailty individuals [45]. Although the study only considered 10 has not been previously investigated as an outcome, other chronic conditions to assess multimorbidity, it supports the studies have demonstrated that multimorbidity patterns have idea of adverse health outcomes being related to the presence a differential impact on healthcare utilisation [36], physical of multiple neuropsychiatric diseases [45]. and cognitive function [24], as well as disability [37] and are Those in the metabolic and sleep disorders pattern had associated with different risk factors [31, 38]. a significantly higher risk of developing frailty within 6, but Cross-sectionally, each clinically specific multimorbidity not 12, years compared with the unspecific pattern. This pat- pattern was significantly associated with physical frailty com- tern is characterised by conditions such as obesity, diabetes pared with the unspecific pattern. The magnitude of the and sleep disorders, which impact physical activity, fatigue, odds ratios suggests that the cardiovascular diseases, anaemia, and thus, frailty [46–48]. A reason for the lack of long- and dementia and psychiatric diseases patterns were most term association with frailty could be that these individuals strongly associated with physical frailty at baseline, compared have developed more severe cardiovascular diseases, which with the unspecific pattern. This could simply be an expres- could lead them to change pattern membership, diluting the sion of the severity and functional impairment caused by the association with incident frailty. The sensory impairments diseases in these patterns, such as dementia, depression and and cancer pattern was significantly associated with frailty cerebrovascular diseases, which have systemic implications development compared with the unspecific pattern, over [39–41]. 12 years. Participants with these patterns likely had less severe 2188 Multimorbidity patterns and risk of frailty in older community-dwelling adults cancers at baseline to be well enough to enrol in SNAC-K, Swedish adults. Persons in multimorbidity patterns charac- which eventually could lead to frailty. terised by cardiovascular and neuropsychiatric diseases are at Understanding how different multimorbidity patterns are increased risk of developing physical frailty. Further research associated with frailty can inform guidelines, prognoses and should investigate the specific mechanisms responsible for preventive strategies for groups at higher frailty risk. Frailty this differential risk between different multimorbidity pat- results in increased healthcare utilisation and its management terns and physical frailty. is costly [8]. An enhanced understanding of the features of older multimorbid persons at risk for frailty can allow Supplementary Data: Supplementary data mentioned in public health authorities to plan community interventions the text are available to subscribers in Age and Ageing online. and health resources accordingly. Generating homogeneous groups of individuals based on multimorbidity patterns also Acknowledgements: We thank the SNAC-K participants has important implications for research and clinical man- and the SNAC-K Group for their collaboration in data agement. Clinicians can provide improved prognoses and collection and management. treatment plans for patients based on their multimorbidity Declaration of Conflicts of Interest: None. pattern, implementing treatment programs such as physical activity to prevent or delay frailty onset. Moreover, this Declaration of Sources of Funding: This work was calls for future research into the biological mechanisms supported by the funders of the Swedish National study responsible for these different health impacts. on Ageing and Care (SNAC): the Ministry of Health and This study is strengthened by its long follow-up, mak- Social Affairs, Sweden; the participating County Councils ing it possible to investigate multimorbidity patterns and and Municipalities; the Swedish Research Council for frailty-onset across several time-points. We took the novel Medicine and Health [VR; 521-2013-8676, 2017-06088, approach of evaluating physical frailty-risk by pattern within 2016-00981]; and the Swedish Research Council for Health, a multimorbid population, rather than comparing multi- Working life and Welfare [Forte; 2016-07175, 2017- morbid participants to a healthy reference group. Measuring 01764]. This work was also supported by the Italian Ministry multimorbidity through patterns, rather than disease count, of Health [PE-2016-02364885]. The funders had no role provides a more nuanced classification of multimorbidity in any step of the present study, from study protocol to and a qualitative dimension that can be useful in clinical manuscript preparation. practice and research. Data regarding multimorbidity and frailty were high-quality, as SNAC-K involves comprehen- sive clinical assessments and participant interviews [49]. Moreover, the rigorous disease classification and advanced References clustering methods provided a unique opportunity to inves- tigate multimorbidity patterns in relation to frailty. The 1. van den Akker M, Buntinx F, Knottnerus J. Comorbidity or sensitivity analyses, involving imputations, also strengthened multimorbidity: what’s in a name? a review of literature. Eur the results. J Gen Pract 1996; 2: 65–70. Nonetheless, there are limitations to consider. Selection 2. Marengoni A, Angleman S, Melis R et al. Aging with mul- bias could have been introduced through the 73% response timorbidity: a systematic review of the literature. Ageing Res rate. Information bias might have occurred, as the exposure Rev 2011; 10: 430–9. and outcome assessments involved self-reported data. To 3. Nunes BP, Flores TR, Mielke GI, Thumé E, Facchini LA. Multimorbidity and mortality in older adults: a systematic limit this, multiple sources (questionnaires, hospital records, review and meta-analysis. Arch Gerontol Geriatr 2016; 67: clinical assessments and proxies for those with cognitive 130–8. impairment) were used. Given the complex and dynamic 4. Librero J, Peiró S, Ordiñana R. Chronic comorbidity and nature of multimorbidity in older adults, it is possible that outcomes of hospital care: length of stay, mortality, and read- participants’ multimorbidity pattern statuses changed since mission at 30 and 365 days. J Clin Epidemiol 1999; 52: baseline. Many in the unspecific pattern would have changed 171–9. to a pattern with a more burdensome disease status, possibly 5. Dumbreck S, Flynn A, Nairn M et al. Drug-disease and drug- underestimating specific patterns’ associations with frailty. drug interactions: systematic examination of recommenda- It is also likely that many older participants died before tions in 12 UK national clinical guidelines. BMJ 2015; 350: they could be captured as frail, possibly underestimating the h949. association. Confounding might have impacted the relation 6. Vetrano DL, Calderón-Larrañaga A, Marengoni A et al. An international perspective on chronic multimorbidity: between the multimorbidity patterns and physical frailty— approaching the elephant in the room. J Gerontol A Biol Sci as such, extensive covariate adjustment was performed. Med Sci 2018; 73: 1350–6. Lastly, the external validity should be critically appraised, 7. Clegg A, Young J, Iliffe S, Rikkert MO, Rockwood K. Frailty as participants were limited to Kungsholmen, a relatively in elderly people. Lancet 2013; 381: 752–62. affluent area of Stockholm. 8. Hoogendijk EO, Afilalo J, Ensrud KE, Kowal P, Onder G, This study found specific multimorbidity patterns to be Fried LP. Frailty: implications for clinical practice and public differentially associated with incident physical frailty in older health. Lancet 2019; 394: 1365–75. 2189 C. Tazzeo et al. 9. Shamliyan T, Talley KM, Ramakrishnan R, Kane RL. Asso- PLoS Med 2018; 15: e1002503. https://doi.org/10.1371/jou ciation of frailty with survival: a systematic literature review. rnal.pmed.1002503. Ageing Res Rev 2013; 12: 719–36. 25. Vetrano DL, Roso-Llorach A, Fernández S et al. Twelve- 10. Kojima G. Frailty as a predictor of future falls among year clinical trajectories of multimorbidity in a population community-dwelling older people: a systematic review and of older adults. Nat Commun 2020; 11: 3223. https://doi.o meta-analysis. J Am Med Dir Assoc 2015; 16: 1027–33. rg/10.1038/s41467-020-16780-x. 11. Hoogendijk EO, Suanet B, Dent E, Deeg DJ, Aartsen MJ. 26. Machón M, Mateo-Abad M, Clerencia-Sierra M et al. Mul- Adverse effects of frailty on social functioning in older adults: timorbidity and functional status in older people: a cluster results from the Longitudinal Aging Study Amsterdam. Matu- analysis. Eur Geriatr Med 2020; 11: 321–32. ritas 2016; 83: 45–50. 27. Santoni G. Data Description. SNAC-K Population Study. 12. Kojima G, Iliffe S, Jivraj S, Walters K. Association between Updated September 26, 2019. https://www.snac-k.se/for- frailty and quality of life among community-dwelling older researchers/data-description/ (1 October 2020, date last people: a systematic review and meta-analysis. J Epidemiol accessed). Community Health 2016; 70: 716–21. 28. Äldrecentrum. Study Plan. SNAC-K Population Study. 13. O’Caoimh R, Sezgin D, O’Donovan MR et al. Prevalence of https://www.snac-k.se/about/study-plan/ (1 October 2020, frailty in 62 countries across the world: a systematic review and date last accessed). meta-analysis of population-level studies. Age Ageing 2021; 29. Calderón-Larrañaga A, Vetrano DL, Onder G et al. Assessing 50: 96–104. and measuring chronic multimorbidity in the older popula- 14. Hanlon P, Nicholl BI, Jani BD, Lee D, McQueenie R, tion: a proposal for its operationalization. J Gerontol A Biol Mair FS. Frailty and pre-frailty in middle-aged and older Sci Med Sci 2017; 72: 1417–23. adults and its association with multimorbidity and mortality: 30. Fried LP, Tangen CM, Walston J et al. Frailty in older adults: a prospective analysis of 493 737 UK Biobank participants. evidence for a phenotype. J Gerontol A Biol Sci Med Sci 2001; Lancet Public Health 2018; 3: e323–32. 56: M146–57. 15. Alvarado BE, Zunzunegui M-V, Béland F, Bamvita J-M. Life 31. Marengoni A, Roso-Llorach A, Vetrano DL et al. Patterns course social and health conditions linked to frailty in Latin of multimorbidity in a population-based cohort of older American older men and women. J Gerontol A Biol Sci Med people: sociodemographic, lifestyle, clinical, and functional Sci 2008; 63: 1399–406. differences. J Gerontol A Biol Sci Med Sci 2020; 75: 798–805. 16. de Souto Barreto P, Greig C, Ferrandez A-M. Detecting 32. Azur MJ, Stuart EA, Frangakis C, Leaf PJ. Multiple imputa- and categorizing frailty status in older adults using a self- tion by chained equations: what is it and how does it work? report screening instrument. Arch Gerontol Geriatr 2012; 54: Int J Methods Psychiatr Res 2011; 20: 40–9. e249–54. 33. Rubin DB. Multiple Imputation for Nonresponse in Surveys. 17. Castell M-V, Sánchez M, Julián R, Queipo R, Martín New York: John Wiley and Sons, 1987. S, Otero Á. Frailty prevalence and slow walking speed 34. Espinoza SE, Jung I, Hazuda H. Lower frailty incidence in in persons age 65 and older: implications for primary older Mexican Americans than in older European Americans: care. BMC Fam Pract 2013; 14: 86. https://doi.o the San Antonio Longitudinal Study of Aging. J Am Geriatr rg/10.1186/1471-2296-14-86. Soc 2010; 58: 2142–8. 18. Wong CH, Weiss D, Sourial N et al. Frailty and its association 35. Jung H-W, Jang I-Y, Lee YS et al. Prevalence of frailty with disability and comorbidity in a community-dwelling and aging-related health conditions in older Koreans in rural sample of seniors in Montreal: a cross-sectional study. Aging communities: a cross-sectional analysis of the Aging Study Clin Exp Res 2010; 22: 54–62. of Pyeongchang Rural Area. J Korean Med Sci 2016; 31: 19. Chang SS, Weiss CO, Xue Q-L, Fried LP. Patterns of comor- 345–52. bid inflammatory diseases in frail older women: the Women’s 36. Juul-Larsen HG, Christensen LD, Bandholm T et al. Patterns Health and Aging Studies I and II. J Gerontol A Biol Sci Med of multimorbidity and differences in healthcare utilization and Sci 2010; 65: 407–13. complexity among acutely hospitalized medical patients (≥65 20. Mendonça N, Kingston A, Yadegarfar M et al. Transitions Years) – a latent class approach. Clin Epidemiol 2020; 12: between frailty states in the very old: the influence of socioe- 245–59. conomic status and multi-morbidity in the Newcastle 85+ 37. Quiñones AR, Markwardt S, Botoseneanu A. Multimorbidity cohort study. Age Ageing 2020; 49: 974–81. combinations and disability in older adults. J Gerontol A 21. Vetrano DL, Palmer K, Marengoni A et al. Frailty and multi- Biol Sci Med Sci 2016; 71: 823–30. morbidity: a systematic review and meta-analysis. J Gerontol 38. Jackson CA, Dobson AJ, Tooth LR, Mishra GD. Lifestyle A Biol Sci Med Sci 2019; 74: 659–66. and socioeconomic determinants of multimorbidity pat- 22. Prados-Torres A, Calderón-Larrañaga A, Hancco-Saavedra J, terns among mid-aged women: a longitudinal study. PLoS Poblador-Plou B, van den Akker M. Multimorbidity pat- One 2016; 11: e0156804. https://doi.org/10.1371/journal. terns: a systematic review. J Clin Epidemiol 2014; 67: pone.0156804. 254–66. 39. Mitchell SL, Teno JM, Kiely DK et al. The clinical course of 23. Calderón-Larrañaga A, Vetrano DL, Ferrucci L et al. Multi- advanced dementia. New Engl J Med 2009; 361: 1529–38. morbidity and functional impairment–bidirectional interplay, 40. American Psychiatric Association. Diagnostic and Statistical synergistic effects and common pathways. J Intern Med 2019; Manual of Mental Disorders. 5th edition. Arlington, VA: 285: 255–71. American Psychiatric Association, 2013. 24. Vetrano DL, Rizzuto D, Calderón-Larrañaga A et al. Trajecto- 41. Pantoni L. Cerebral small vessel disease: from pathogenesis ries of functional decline in older adults with neuropsychiatric and clinical characteristics to therapeutic challenges. Lancet and cardiovascular multimorbidity: a Swedish cohort study. Neurol 2010; 9: 689–701. 2190 Multimorbidity patterns and risk of frailty in older community-dwelling adults 42. Feigenbaum MS, Welsch MA, Mitchell M, Vincent K, Braith 46. Forhan M, Gill SV. Obesity, functional mobility and quality RW, Pepine CJ. Contracted plasma and blood volume in of life. Best Pract Res Clin Endocrinol Metab 2013; 27: chronic heart failure. J Am Coll Cardiol 2000; 35: 51–5. 129–37. 43. Kokkinos PF, Choucair W, Graves P, Papademetriou V, 47. Sinclair AJ, Rodriguez-Mañas L. Diabetes and frailty: two Ellahham S. Chronic heart failure and exercise. Am Heart J converging conditions? Can J Diabetes 2016; 40: 77–83. 2000; 140: 21–8. 48. Buysse DJ. Insomnia. JAMA 2013; 309: 706–16. 44. van Alphen HJ, Volkers KM, Blankevoort CG, Scherder EJ, 49. Lagergren M, Fratiglioni L, Hallberg IR et al. A longitudinal Hortobágyi T, van Heuvelen MJ. Older adults with dementia study integrating population, care and social services data. The are sedentary for most of the day. PLoS One 2016; 11: Swedish National study on Aging and Care (SNAC). Aging e0152457. https://doi.org/10.1371/journal.pone.0152457. Clin Exp Res 2004; 16: 158–68. 45. Nguyen QD, Wu C, Odden MC, Kim DH. Multimorbidity patterns, frailty, and survival in community-dwelling older adults. J Gerontol A Biol Sci Med Sci 2019; 74: 1265–70. Received 1 December 2020; editorial decision 6 May 2021 2191