Advertisement

Transitions in Frailty and 4-Year Mortality Risk in Taiwan Longitudinal Study on Aging

  • An-Chun Hwang
    Affiliations
    Center for Geriatrics and Gerontology, Taipei Veterans General Hospital, Taipei, Taiwan

    Department of Geriatric Medicine, National Yang Ming Chiao Tung University, School of Medicine, Taipei, Taiwan

    Institute of Public Health, National Yang Ming Chiao Tung University, Taipei, Taiwan

    Center for Healthy Longevity and Aging Sciences, National Yang Ming Chiao Tung University, Taipei, Taiwan
    Search for articles by this author
  • Liang-Yu Chen
    Affiliations
    Center for Geriatrics and Gerontology, Taipei Veterans General Hospital, Taipei, Taiwan

    Department of Geriatric Medicine, National Yang Ming Chiao Tung University, School of Medicine, Taipei, Taiwan

    Center for Healthy Longevity and Aging Sciences, National Yang Ming Chiao Tung University, Taipei, Taiwan
    Search for articles by this author
  • Ting-Ching Tang
    Affiliations
    Center for Geriatrics and Gerontology, Taipei Veterans General Hospital, Taipei, Taiwan

    Department of Geriatric Medicine, National Yang Ming Chiao Tung University, School of Medicine, Taipei, Taiwan
    Search for articles by this author
  • Li-Ning Peng
    Affiliations
    Center for Geriatrics and Gerontology, Taipei Veterans General Hospital, Taipei, Taiwan

    Department of Geriatric Medicine, National Yang Ming Chiao Tung University, School of Medicine, Taipei, Taiwan

    Center for Healthy Longevity and Aging Sciences, National Yang Ming Chiao Tung University, Taipei, Taiwan
    Search for articles by this author
  • Ming-Hsien Lin
    Affiliations
    Center for Geriatrics and Gerontology, Taipei Veterans General Hospital, Taipei, Taiwan

    Department of Geriatric Medicine, National Yang Ming Chiao Tung University, School of Medicine, Taipei, Taiwan

    Center for Healthy Longevity and Aging Sciences, National Yang Ming Chiao Tung University, Taipei, Taiwan
    Search for articles by this author
  • Yiing-Jenq Chou
    Correspondence
    Yiing-Jenq Chou, MD, PhD, Institute of Public Health, National Yang Ming Chiao Tung University, No.115, Sec. 2, Li-Nong Street, Taipei 112, Taiwan.
    Affiliations
    Institute of Public Health, National Yang Ming Chiao Tung University, Taipei, Taiwan

    Office of the Deputy Superintendent, National Yang Ming Chiao Tung University Hospital, Yilan County, Taiwan
    Search for articles by this author
  • Fei-Yuan Hsiao
    Affiliations
    Graduate Institute of Clinical Pharmacy, College of Medicine, National Taiwan University, Taipei, Taiwan

    School of Pharmacy, College of Medicine, National Taiwan University, Taipei, Taiwan

    Department of Pharmacy, National Taiwan University Hospital, Taipei, Taiwan
    Search for articles by this author
  • Liang-Kung Chen
    Correspondence
    Address correspondence to Liang-Kung Chen, MD, PhD, Center for Geriatrics and Gerontology, Taipei Veterans General Hospital, No. 201, Sec. 2, Shih-Pai Road, Taipei, Taiwan.
    Affiliations
    Center for Geriatrics and Gerontology, Taipei Veterans General Hospital, Taipei, Taiwan

    Department of Geriatric Medicine, National Yang Ming Chiao Tung University, School of Medicine, Taipei, Taiwan

    Center for Healthy Longevity and Aging Sciences, National Yang Ming Chiao Tung University, Taipei, Taiwan

    Taipei Municipal Gan-Dau Hospital (Managed by Taipei Veterans General Hospital), Taipei, Taiwan
    Search for articles by this author
Published:November 09, 2022DOI:https://doi.org/10.1016/j.jamda.2022.10.005

      Abstract

      Objectives

      To explore the associations of (1) the frailty phenotype or frailty index transition with cause-specific mortality, and (2) different combinations of transition in frailty phenotype and frailty index with all-cause mortality.

      Design

      Retrospective cohort study.

      Setting and Participants

      Data from 3529 respondents aged >50 years who completed the 1999 and 2003 surveys of the Taiwan Longitudinal Study on Aging were analyzed.

      Methods

      Cox regression and subdistribution hazard models were constructed to investigate frailty phenotype or frailty index transitions (by categories of frailty phenotype, absolute and percentage changes in frailty index, and combined categories of the 2 measurements) and subsequent 4-year all-cause and cause-specific mortality, respectively.

      Results

      Among the frailty phenotype transition groups, the improved frailty group had overall mortality risk comparable to that of the maintained robustness/prefrailty group [hazard ratio (HR): 0.9; 95% CI: 0.7–1.2] and lower risk of mortality due to organ failure (HR: 0.4; 95% CI: 0.2–0.8; P = .015), whereas the worsened frailty group had the highest risk of all-cause mortality and death from infection, malignancy, cardiometabolic/cerebrovascular diseases, and other causes (HR: 1.8–3.7; all P < .03). The rapidly increased frailty index group had significantly higher all-cause and every cause-specific mortality than the decreased frailty index group (HR: 1.8–7.7; all P < .05). When frailty phenotype and frailty index transition groups were combined, participants with worsened frailty/rapidly increased frailty index had increased risk under the same frailty index/frailty phenotype transition condition, particularly for large changes in each factor (HR: 1.5–2.2; P < .01 for worsened frailty; 1.7–4.5, P < .03 for rapidly increased frailty index).

      Conclusions and Implications

      We found that considering both frailty phenotype and frailty index provided best mortality prediction. These associations were independent of baseline frailty status and comorbidities. Nevertheless, even capturing transitions in frailty phenotype or frailty index only can provide good mortality prediction, which supported adopting these approaches in different clinical settings.

      Keywords

      Frailty is characterized by declines in physiological reserves across different systems and increased vulnerability to stressors during the process of aging.
      • Clegg A.
      • Young J.
      • Iliffe S.
      • et al.
      Frailty in elderly people.
      As a surrogate of accelerated aging, frailty has been shown to be an important confounder for the treatment of chronic diseases in older people
      • Draznin B.
      • Aroda V.R.
      • et al.
      American Diabetes Association Professional Practice Committee
      13. Older Adults: Standards of Medical Care in Diabetes-2022.
      ,
      • Williams B.
      • Mancia G.
      • Spiering W.
      • et al.
      2018 ESC/ESH Guidelines for the management of arterial hypertension.
      and independently predictive of unfavorable outcomes, including falls,
      • Kojima G.
      Frailty as a predictor of future falls among community-dwelling older people: a systematic review and meta-analysis.
      reduced quality of life,
      • Kojima G.
      • Iliffe S.
      • Morris R.W.
      • et al.
      Frailty predicts trajectories of quality of life over time among British community-dwelling older people.
      institutionalization, and mortality.
      • Rockwood K.
      • Mitnitski A.
      • Song X.
      • et al.
      Long-term risks of death and institutionalization of elderly people in relation to deficit accumulation at age 70.
      ,
      • Fried L.P.
      • Tangen C.M.
      • Walston J.
      • et al.
      Frailty in older adults: evidence for a phenotype.
      Multiple studies have investigated the association between baseline frailty status and all-cause mortality.
      • Chang S.F.
      • Lin P.L.
      Frail phenotype and mortality prediction: a systematic review and meta-analysis of prospective cohort studies.
      ,
      • Kojima G.
      • Iliffe S.
      • Walters K.
      Frailty index as a predictor of mortality: a systematic review and meta-analysis.
      Nevertheless, frailty status fluctuates, and the impact of a one-time frailty assessment is more predictive of short-term than long-term mortality.
      • Thompson M.Q.
      • Theou O.
      • Tucker G.R.
      • et al.
      Recurrent measurement of frailty is important for mortality prediction: findings from the North West Adelaide Health Study.
      It has been indicated that frailty changes, whether measured by frailty phenotype
      • Kulmala J.
      • Nykanen I.
      • Hartikainen S.
      Frailty as a predictor of all-cause mortality in older men and women.
      ,
      • Wang M.C.
      • Li T.C.
      • Li C.I.
      • et al.
      Frailty, transition in frailty status and all-cause mortality in older adults of a Taichung community-based population.
      or the frailty index,
      • Stolz E.
      • Hoogendijk E.O.
      • Mayerl H.
      • et al.
      Frailty changes predict mortality in four longitudinal studies of aging.
      ,
      • Shi S.M.
      • Olivieri-Mui B.
      • McCarthy E.P.
      • et al.
      Changes in a frailty index and association with mortality.
      the 2 most commonly adopted frailty assessment models, are predictive of long-term mortality despite baseline frailty status. Although comparisons between the 2 measurements at baseline and subsequent mortality have been investigated,
      • Rockwood K.
      • Andrew M.
      • Mitnitski A.
      A comparison of two approaches to measuring frailty in elderly people.
      longitudinal comparisons remain to be explored. It is particularly of interest for the group with a discrepant transition between the 2 measurements (eg, a favorable frailty index transition + an unfavorable frailty phenotype transition).
      In addition to the association with all-cause mortality, the association between frailty status and cause-specific mortality has attracted increasing research interest. Recent studies have suggested that the risk of cause-specific mortality due to cardiometabolic diseases, cancer, and infection may increase with increasing severity of the baseline frailty phenotype or frailty index.
      • Lohman M.C.
      • Sonnega A.J.
      • Resciniti N.V.
      • et al.
      Frailty phenotype and cause-specific mortality in the United States.
      • Grabovac I.
      • Haider S.
      • Mogg C.
      • et al.
      Frailty status predicts all-cause and cause-specific mortality in community dwelling older adults.
      • Fan J.
      • Yu C.
      • Guo Y.
      • et al.
      Frailty index and all-cause and cause-specific mortality in Chinese adults: a prospective cohort study.
      The association between longitudinal frailty transition and cause-specific mortality has been rarely investigated and may provide more information for long-term risk prediction.
      In this context, we analyzed data from a nationally representative cohort in Taiwan, with the specific aims of exploring the association between (1) frailty phenotype and frailty index transitions and subsequent 4-year all-cause and cause-specific mortality, and (2) different combinations of frailty phenotype/frailty index transitions and all-cause mortality.

      Methods

      The Taiwan Longitudinal Study on Aging (TLSA) is a nationally representative cohort study that includes community-dwelling residents older than 50 years and has been conducted by the Taiwan Health Promotion Administration since 1989; the respondents are followed every 3 to 4 years.
      Health Promotion Administration, Ministry of Health and Welfare, Taiwan, Taiwan Longitudinal Study on Aging (TLSA).
      The study aims to address potential impacts on the economic, social, and medical aspects of population aging. We extracted data from participants of the 1999 TLSA survey (n = 4440) and excluded those who had died or were lost to follow-up before the 2003 TLSA survey (n = 911). Those who participated in both waves were recruited for final analysis to describe frailty phenotype and frailty index transitions (n = 3529).
      The study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the Institutional Review Board (No. 2021-05-023CC).

      Construction of Frailty Phenotype and Frailty Index

      The definitions of frailty phenotype and frailty index have been previously described in detail,
      • Hwang A.C.
      • Lee W.J.
      • Huang N.
      • et al.
      Longitudinal changes of frailty in 8 years: comparisons between physical frailty and frailty index.
      which were mostly based on self-reported questionnaire items. In brief, the 5 components of frailty phenotype were modified according to the original Fried criteria
      • Fried L.P.
      • Tangen C.M.
      • Walston J.
      • et al.
      Frailty in older adults: evidence for a phenotype.
      : (1) Weakness was defined as difficulty in picking up or twisting objects using the fingers. (2) Slowness was defined as difficulty walking 200 to 300 m. (3) Weight loss was defined as a body mass index ≤18.5. (4) Exhaustion was defined as self-reported answers of “I felt everything I did was an effort” or “I could not get going” >2 days/week for either answer on the Center for Epidemiological Studies-Depression Scale (CES-D) questionnaire. (5) Inactivity was defined as a sum of the weighted score ≤3 for men or ≤2 for women calculated from the frequency and intensity of leisure time activities.
      • Chen C.Y.
      • Wu S.C.
      • Chen L.J.
      • et al.
      The prevalence of subjective frailty and factors associated with frailty in Taiwan.
      Frailty index was calculated according to a standard procedure.
      • Searle S.D.
      • Mitnitski A.
      • Gahbauer E.A.
      • et al.
      A standard procedure for creating a frailty index.
      Seventy-two variables were selected, including the following: (1) activities of daily living (ADLs), instrumental ADLs (IADLs), and mobility (22 variables); (2) comorbidities and health status (17 variables), and 13 comorbidities were ascertained by asking “have you ever been diagnosed with the following diseases by a physician?”; (3) psychological status (10 variables); (4) cognitive function (10 variables); (5) life satisfaction (6 variables); (6) stress (4 variables) and sensory domain (3 variables) (Supplementary Table 1).
      • Hwang A.C.
      • Lee W.J.
      • Huang N.
      • et al.
      Longitudinal changes of frailty in 8 years: comparisons between physical frailty and frailty index.
      For each individual, the frailty index was calculated as the sum of deficits divided by the maximum possible score.

      Definitions of Frailty Phenotype and Frailty Index Transitions

      Four approaches in capturing transitions in frailty [by categories of phenotype, categories (absolute changes) of frailty index, percentage changes in frailty index, and combined categories of frailty phenotype and frailty index]. The rationale we adopt these approaches is to test the feasibility of these measures in capturing transitions in frailty and their associations with risk of mortality.
      Transitions in frailty phenotype and frailty index were calculated as the differences between values measured in 1999 and 2003. A positive value suggested worsening of frailty status. We categorized transition in frailty phenotype into 4 groups: (1) worsened frailty (frailty phenotype20031999 >0), (2) improved frailty (frailty phenotype2003–1999 <0), (3) maintained robustness/prefrailty (frailty phenotype2003–1999 = 0 and frailty phenotype1999 = 0–2), and (4) maintained frailty (frailty phenotype2003–1999 = 0 and frailty phenotype1999 = 3–5).
      Because of the lack of consensus regarding the definition of frailty index transition, we adopted the minimally clinically significant change of 0.03/year proposed by Jang et al.
      • Jang I.Y.
      • Jung H.W.
      • Lee H.Y.
      • et al.
      Evaluation of clinically meaningful changes in measures of frailty.
      Thus, a frailty index increase of more than 0.12 in 4 years was considered a rapid increase, and frailty index transition was classified as the following categories: (1) decreased frailty index (frailty index2003–1999 < 0); (2) moderately increased frailty index (0 ≤ frailty index2003–1999 <0.12); and (3) rapidly increased frailty index (frailty index2003–1999 ≥0.12). In addition, we categorized the frailty index by percentage change: (1) decrease in the frailty index percentage: <0%; (2) moderate increase in the frailty index percentage: increase <50%; and (3) rapid increase in the frailty index percentage: increase ≥50% to check the robustness of the association between frailty index transition and mortality.
      We further combined transitions in frailty phenotype (worsened frailty, improved frailty, maintained robustness/prefrailty, and maintained frailty) and transition in frailty index (decreased, moderately increased, and rapidly increased frailty index) to create 12 subcategories of transitions in both frailty phenotype and frailty index.

      Mortality

      Mortality was assessed from the 2003 TLSA survey until Dec 2007. The main International Classification of Diseases, ninth revision (ICD-9), codes for death as well as year and month of death were collected from the death certificate reporting system of the Ministry of Health and Welfare, Taiwan. We considered 5 causes of death in our analysis: infectious diseases, malignancy, cardiometabolic and cerebrovascular diseases, organ failure (including heart, lung, kidney, and liver), and others. The corresponding ICD-9 codes for each cause of death are presented in Supplementary Table 2.

      Statistical Analysis

      Descriptive statistics are presented as mean ± SDs for continuous variables or numbers (percentages) for categorical variables. Cox proportional hazard regression models were used to evaluate the relationship between frailty transition and subsequent 4-year all-cause mortality. Model 1 was adjusted for age and sex; model 2 was adjusted for age, sex, education years, marital status, alcohol consumption, smoking, and total number of comorbidities; and model 3 was further adjusted for baseline frailty phenotype or frailty index in addition to the variables in model 2 (Table 1). The transition groups of frailty phenotype and frailty index were combined in a fully adjusted Cox regression model to predict the mortality risk.
      Table 1Frailty Phenotype and Frailty Index Transitions and 4-Year All-Cause Mortality
      Model 1
      Model 1: adjusted for age, sex.
      Model 2
      Model 2: adjusted for age, sex, education years, marital status, alcohol, smoke, number of comorbidities.
      Model 3
      Model 3: adjusted for age, sex, education years, marital status, alcohol, smoke, number of comorbidities, baseline frailty phenotype or frailty index.
      HR (95% CI)P valueHR (95% CI)P valueHR (95% CI)P value
      Total655/3529 (18.6)
      Frailty phenotype transition
       Maintained robustness/prefrailty197/1729 (11.4)RefRefRef
       Worsened frailty296/1065 (27.8)2.3 (1.9, 2.8)<.0012.1 (1.7, 2.5)<.0012.0 (1.7, 2.5)<.001
       Improved frailty124/667 (18.6)1.6 (1.3, 2.0).0011.5 (1.2, 1.9).0010.9 (0.7, 1.2).472
       Maintained frailty38/68 (55.9)4.2 (3.0, 6.0)<.0013.4 (2.4, 5.0)<.0011.5 (1.0, 2.3).074
      Frailty index transition defined by absolute change
       Decreased frailty index168/1304 (12.9)RefRefRef
       Moderately increased frailty index (increase <0.12)248/1664 (14.9)1.1 (0.9–1.3).5371.0 (0.8–1.2).8371.3 (1.0–1.6).020
       Rapidly increased frailty index (increase ≥0.12)239/561 (42.6)2.8 (2.3–3.5)<.0012.6 (2.1–3.2)<.0013.3 (2.7–4.2)<.001
      Frailty index transition defined by percentage change
       Decreased frailty index168/1304 (12.9)RefRefRef
       Moderately increased frailty index (increase <50%)175/997 (17.6)1.2 (1.0–1.5).0461.1 (0.9–1.3).5151.3 (1.0–1.6).020
       Rapidly increased frailty index (increase ≥50%)312/1228 (25.4)1.7 (1.4–2.1)<.0011.6 (1.3–1.9)<.0012.6 (2.1–3.3)<.001
      Model 1: adjusted for age, sex.
      Model 2: adjusted for age, sex, education years, marital status, alcohol, smoke, number of comorbidities.
      Model 3: adjusted for age, sex, education years, marital status, alcohol, smoke, number of comorbidities, baseline frailty phenotype or frailty index.
      We used Fine and Gray subdistribution hazard (SDH) models to calculate cause-specific mortality in the different frailty transition groups. Death by any cause other than the cause of interest was defined as a competing event, and separate models were fit for each of the 5 causes of death. SDH models for death from infection and other causes included the same variables as the fully adjusted Cox proportional hazard model, and we performed further adjustment for related baseline comorbidities of malignancy, cardiometabolic, and organ failure. All statistical analyses were performed using SAS software 9.4 (SAS Institute Inc) and SPSS, version 24.0 (IBM Corp), and a P value (2-tailed) of less than 0.05 was considered statistically significant.

      Results

      In total, 3529 participants who completed the 1999 and 2003 waves were enrolled in the final analysis (mean age 67.6 ± 8.6 years in 1999, 51.5% male). Totals of 2041 (57.8%), 1263 (35.8%), and 225 (6.4%) participants were considered robust, prefrail, and frail, respectively, as defined by the frailty phenotype in 1999; the mean frailty index was 0.15 (SD 0.11). After 4 years, the prevalence rates of all comorbidities and the degree of frailty, either defined by frailty phenotype or frailty index, increased (Supplementary Table 3).

      Frailty Phenotype or Frailty Index Transition and Subsequent 4-Year All-Cause Mortality

      As of December 31, 2007, 2871 (81.4%) participants were alive, 3 (0.1%) had moved to other countries, and 655 (18.6%) had died. Regarding frailty phenotype transition and all-cause mortality, in comparison with the maintained robustness/prefrailty group, the maintained frailty group had the highest overall mortality, followed by the worsened frailty group, after adjustment for baseline characteristics (model 2) (all P ≤ .001). In the fully adjusted model, which took the baseline frailty phenotype score into consideration, the improved frailty group had a risk similar to the maintained robustness/prefrailty group (HR: 0.9; 95% CI: 0.7–1.2), and the worsened frailty group had the highest risk (HR: 2.0; 95% CI: 1.7–2.5; P < .001).
      Regarding frailty index transition (defined by absolute change) and all-cause mortality, the dose–response relationship became significant after adjustment for baseline frailty index. The HR of the moderately increased and rapidly increased frailty index groups were 1.3 (95% CI: 1.0–1.6; P = .02) and 3.3 (95% CI: 2.7–4.2; P < .001), respectively. The results were similar when frailty index transition was defined as the percentage change (Table 1, Figure 1, A and B ).
      Figure thumbnail gr1
      Fig. 1(A) Frailty phenotype transition and all-cause mortality. (B) Frailty index transition and all-cause mortality.

      Frailty Phenotype or Frailty Index Transition and Subsequent 4-Year Cause-specific Mortality

      Among 655 deaths, 72 (11.0%) were due to infectious diseases, 168 (25.6%) were due to malignancy, 189 (28.9%) were due to cardiometabolic/cerebrovascular diseases, 130 (19.8%) were due to organ failure, and 96 (14.7%) were due to other causes.
      Regarding frailty phenotype transition and cause-specific mortality, in comparison with the maintained robustness/prefrailty group, the worsened frailty group had significantly higher risks of death from infection (HR: 3.7; 95% CI: 1.9–7.4; P < .001), malignancy (HR: 2.0; 95% CI: 1.4–3.0; P < .001), cardiometabolic/cerebrovascular diseases (HR: 1.8; 95% CI 1.2–2.7; P = .003), and other causes (HR: 1.8; 95% CI 1.1–3.1; P = .027), whereas the improved frailty group had risks comparable risks to the maintained robustness/prefrailty group (HR ranged from 1.1–1.4; P > .2) for these 4 causes of death and a lower risk of death from organ failure (HR: 0.4; 95% CI 0.2–0.9; P = .017). The maintained frailty group had borderline increased mortality risks from infection (HR: 2.9; 95% CI 0.9–9.8; P = .084) and malignancy (HR: 2.8; 95 CI 0.9–8.9, P = .086) (Figure 2).
      Figure thumbnail gr2
      Fig. 2Frailty phenotype transition and cause-specific mortality.
      Similarly, the rapidly increased frailty index group had remarkably higher risks of each cause of death than the decreased frailty index group (HR: 7.7; 95% CI 3.4–14.6 for infection; P < .001; HR: 1.8–2.5, CI ranged from 1.1 to 4.3, P < .03 for other 4 causes of death). In contrast, a moderate increase in the frailty index was predictive of a higher risk of infection-related mortality (HR: 3.3; 95% CI 1.5–7.4; P = .004) than mortality due to other causes of death in our analysis (Figure 3). The results were generally consistent when frailty index transition was defined as percentage change (Supplementary Table 4). These associations were independent of baseline comorbidities and frailty phenotype/frailty index.
      Figure thumbnail gr3
      Fig. 3Frailty index transition and cause-specific mortality.

      Combined Frailty Index/Frailty Phenotype Transitions and All-Cause Mortality

      Participants with decreased frailty index had a statistically similar mortality risk despite frailty phenotype transition, and the worsened frailty group had higher risk than the maintained robustness/prefrailty group when the frailty index increased moderately or rapidly (HR: 1.5; 95% CI 1.1–2.1; P = .006; HR: 2.2; 95% CI 1.4–3.3, P < .001 for moderately/rapidly increased frailty index respectively) (Figure 4A).
      Figure thumbnail gr4
      Fig. 4(A) Frailty phenotype transition and all-cause mortality by frailty index. (B) Frailty index transition and all-cause mortality by frailty phenotype.
      On the other hand, compared with the decreased frailty index group, the rapidly increased frailty index group had significantly higher risk among the different frailty phenotype transitions (HR: 1.7; 95% CI: 1.1–2.8; P = .027 for the maintained robustness/prefrailty group; HR: 4.5; 95% CI: 2.9–7.0; P < .001 for the worsened frailty group; HR: 3.0; 95% CI: 1.6–5.7; P < .001 for the improved frailty group; HR: 2.4; 95% CI: 0.9–6.6; P = .089 for the maintained frailty group). The dose–response relationship between frailty index transition and mortality was more significant in the worsened and improved frailty groups (Figure 4B).
      The results were generally similar when frailty index transition was defined as percentage change, whereas the predictive ability of frailty index transition was less significant in the maintained robustness/prefrailty and maintained frailty groups (Supplementary Tables 5 and 6).

      Discussion

      In our study, we calculated and evaluated 4 approaches in capturing transitions in frailty [by categories of phenotype, categories (absolute changes) of frailty index, percentage changes in frailty index, and combined categories of phenotype and frailty index (12 subcategories)] on risk of mortality. We found that considering both frailty phenotype and frailty index provided the best predictability of mortality risk. Nevertheless, even capturing transitions in frailty phenotype or frailty index only can provide good predictability of mortality risk, which provided feasibility in adopting the approaches in different clinical settings.
      Published studies have shown that an increased mortality risk is associated with increasing frailty phenotype levels (eg, robust to prefrail or frail, or prefrail to frail), whereas the impact of frailty phenotype improvement is controversial.
      • Kulmala J.
      • Nykanen I.
      • Hartikainen S.
      Frailty as a predictor of all-cause mortality in older men and women.
      ,
      • Li C.Y.
      • Al Snih S.
      • Karmarkar A.
      • et al.
      Early frailty transition predicts 15-year mortality among nondisabled older Mexican Americans.
      We defined frailty phenotype change with continuous instead of categorical variables, and the effect of frailty phenotype change on mortality risk was confounded but not modified by baseline measurement. The interaction term baseline frailty phenotype (robust, prefrail, frail) frailty phenotype change was not statistically significant in the fully adjusted model (Supplementary Table 7). The results were in accordance with those by Xue et al.
      • Xue Q.L.
      • Bandeen-Roche K.
      • Tian J.
      • et al.
      Progression of physical frailty and the risk of all-cause mortality: is there a point of no return?.
      ; among frail participants, those fulfilling 5 frailty phenotype criteria had a significantly higher 1-year mortality risk than those fulfilling 3 or 4 criteria. A meta-analysis demonstrated that each 0.01 increase in the baseline frailty index was associated with a 4% increase in mortality risk,
      • Kojima G.
      • Iliffe S.
      • Walters K.
      Frailty index as a predictor of mortality: a systematic review and meta-analysis.
      and our study extends the findings longitudinally: the adjusted mortality risk increased by 1.04 (95% CI: 1.04–1.05; P < .001) for each 0.01 increase in the frailty index over 4 years.
      Previous studies have demonstrated that the mortality risk from cancer, cardiovascular diseases, and respiratory disease increased linearly with severity of frailty phenotype at baseline.
      • Lohman M.C.
      • Sonnega A.J.
      • Resciniti N.V.
      • et al.
      Frailty phenotype and cause-specific mortality in the United States.
      ,
      • Grabovac I.
      • Haider S.
      • Mogg C.
      • et al.
      Frailty status predicts all-cause and cause-specific mortality in community dwelling older adults.
      In contrast, an incremental increase in the baseline frailty index, defined by either absolute or percentage change, was associated with excessive risks of cardiovascular- and respiratory-related mortality, whereas the association with cancer-related mortality was less conclusive.
      • Fan J.
      • Yu C.
      • Guo Y.
      • et al.
      Frailty index and all-cause and cause-specific mortality in Chinese adults: a prospective cohort study.
      ,
      • Li X.
      • Ploner A.
      • Karlsson I.K.
      • et al.
      The frailty index is a predictor of cause-specific mortality independent of familial effects from midlife onwards: a large cohort study.
      To the best of our knowledge, very few studies have investigated the association between frailty transition and cause-specific mortality. In our analysis, among the 4 frailty phenotype transition groups, the worsened frailty group had the highest mortality risk from most causes except organ failure. In comparison with other causes of death, participants with major organ failure may have earlier and gradual function decline and higher risk of drop-out from the survey, which may lead to underestimation of the impact of worsened frailty. The maintained frailty group had a borderline increased risk of death from infectious diseases and malignancy, which was congruent with a recent study reporting an increased risk of cancer-related mortality in women with sustained or worsened frailty.
      • Cespedes Feliciano E.M.
      • Hohensee C.
      • Rosko A.E.
      • et al.
      Association of prediagnostic frailty, change in frailty status, and mortality after cancer diagnosis in the Women's Health Initiative.
      The risks in the improved frailty group were comparable to those in the maintained robustness/prefrailty group in most causes of death. The lower risk of organ failure–related death, should be interpreted with caution, which may be influenced by the fluctuation of function before death,
      • Lynn J.
      Perspectives on care at the close of life. Serving patients who may die soon and their families: the role of hospice and other services.
      and there may be unmeasured confounders. Nevertheless, the results highlight the potential importance of frailty and physical performance in mortality prediction in these patients,
      • Yang X.
      • Lupon J.
      • Vidan M.T.
      • et al.
      Impact of frailty on mortality and hospitalization in chronic heart failure: a systematic review and meta-analysis.
      • Puhan M.A.
      • Siebeling L.
      • Zoller M.
      • et al.
      Simple functional performance tests and mortality in COPD.
      • Bao Y.
      • Dalrymple L.
      • Chertow G.M.
      • et al.
      Frailty, dialysis initiation, and mortality in end-stage renal disease.
      although further longitudinal or interventional studies of individual disease are warranted.
      The rapidly increased frailty index group had consistently higher death risk from different causes in our analysis. Nevertheless, the dose–response relationship between frailty index transition and mortality was not evident in death other than infectious diseases. This result implies heterogeneity in the moderately increased frailty index group, whether defined by absolute or percentage change. Shi et al.
      • Shi S.M.
      • Olivieri-Mui B.
      • McCarthy E.P.
      • et al.
      Changes in a frailty index and association with mortality.
      did not observe an impact of a large frailty index increase on mortality in those with a low baseline frailty index, possibly because of adequate reserve in this group. In our analysis, the moderately increased frailty index group had the lowest baseline frailty index compared with the other 2 groups, which may partially explain the results (baseline frailty index: 0.17 ± 0.12 in the improved frailty index group, 0.12 ± 0.10 in the moderately increased frailty index group, and 0.17 ± 0.11 in the rapidly increased frailty index group).
      The frailty phenotype was defined by 5 fixed components and focused on physical frailty.
      • Fried L.P.
      • Tangen C.M.
      • Walston J.
      • et al.
      Frailty in older adults: evidence for a phenotype.
      The underlying mechanism can be possibly attributed to multisystem dysregulation, which can occur independent of comorbidities.
      • Fried L.P.
      • Cohen A.A.
      • Xue Q.L.
      • et al.
      The physical frailty syndrome as a transition from homeostatic symphony to cacophony.
      Frailty defined by phenotype was considered a pre-disability syndrome and predictive value of adverse outcomes may be restricted to nondisabled population and the categorical nature may facilitate clinical interpretation.
      • Cesari M.
      • Gambassi G.
      • van Kan G.A.
      • et al.
      The frailty phenotype and the frailty index: different instruments for different purposes.
      On the other hand, the main concept of frailty index was deficit accumulation, which encompasses symptoms and signs, diseases, and physical, psychosocial function parameters, and so forth.
      • Rockwood K.
      • Song X.
      • MacKnight C.
      • et al.
      A global clinical measure of fitness and frailty in elderly people.
      Users can construct their own frailty index by standard procedure with similar prediction power, and the results were applicable independent of function status.
      • Rockwood K.
      • Mitnitski A.
      • Song X.
      • et al.
      Long-term risks of death and institutionalization of elderly people in relation to deficit accumulation at age 70.
      Difference among frailty phenotype and frailty index assessment tools by 1-time measurement have been compared: the population selected was related but distinct, and the superiority of mortality prediction was heterogeneous.
      • Theou O.
      • Brothers T.D.
      • Mitnitski A.
      • et al.
      Operationalization of frailty using eight commonly used scales and comparison of their ability to predict all-cause mortality.
      ,
      • Woo J.
      • Leung J.
      • Morley J.E.
      Comparison of frailty indicators based on clinical phenotype and the multiple deficit approach in predicting mortality and physical limitation.
      In our analysis, when transition groups by the 2 measurements were combined, the mortality risk increased with the degree of change in the frailty index among the different frailty phenotype transition groups. The dose–response relationships were more significant in the worsened and improved frailty groups. These results are generally consistent with those from our work investigating the relationship between frailty phenotype/frailty index transition and disability.
      • Hwang A.C.
      • Lee W.J.
      • Huang N.
      • et al.
      Longitudinal changes of frailty in 8 years: comparisons between physical frailty and frailty index.
      It also supports the findings that mortality risk increases with higher frailty index among participants with the same frailty phenotype level from a 1-time measurement.
      • Rockwood K.
      • Andrew M.
      • Mitnitski A.
      A comparison of two approaches to measuring frailty in elderly people.
      On the other hand, worsened frailty contributed to the increased risk compared with the other frailty phenotype transition groups among participants with moderately or rapidly increased frailty index. The previously mentioned findings suggested that frailty phenotype and frailty index transitions may be complementary in mortality prediction, especially for those with a large change in either one of these measurements.
      Despite the extensive effort that went into this research, there were several limitations. First, participants who were lost to follow-up before the 2003 survey were not included. The exclusion of these participants, with possibly higher frailty levels or disease severity, might result in low case number in some of the transition groups (especially the maintained frailty group), and underestimate the impact of unfavorable frailty change on mortality. Second, as this is a “survey-based” study in community-dwelling adults, the construction of frailty measurements was mostly defined by self-report items. However, their constructions were based on one of our previously published studies
      • Hwang A.C.
      • Lee W.J.
      • Huang N.
      • et al.
      Longitudinal changes of frailty in 8 years: comparisons between physical frailty and frailty index.
      and other studies
      • Chen C.Y.
      • Wu S.C.
      • Chen L.J.
      • et al.
      The prevalence of subjective frailty and factors associated with frailty in Taiwan.
      ,
      • Searle S.D.
      • Mitnitski A.
      • Gahbauer E.A.
      • et al.
      A standard procedure for creating a frailty index.
      to increase the validity of these measures. Nevertheless, our results could be more robust if some variables could be captured from objective measures such as those documented in the medical record. Third, the frailty status was measured twice, with a 4-year interval, which may be inadequate to capture the change in a timely manner, and whether mortality prediction improves with the frequency of measurement deserves further investigation. Fourth, the lack of consensus regarding the definition of a “moderate increase in the frailty index” may have contributed to the nonsignificant association with most cause-specific deaths. Studies investigating clinically meaningful cutoffs for different outcomes (eg, mortality, disability) may be warranted.

      Conclusions and Implications

      In this nationally representative cohort in Taiwan, middle-aged and older adults with worsened frailty had the highest 4-year overall mortality and mortality due to infection, malignancy, cardiometabolic/cerebrovascular diseases, and other causes among the 4 different frailty phenotype transition groups. The risk of the improved frailty group was generally comparable to that of the maintained robustness/prefrailty group, and the risk of death from organ failure was even lower. On the other hand, in comparison with the decreased frailty index group, the rapidly increased frailty index group had the highest all-cause and cause-specific mortality. These associations were independent of baseline frailty status and comorbidities. When considering both frailty phenotype and frailty index, it provided best predictability of mortality risk in comparison to either frailty phenotype or frailty index. Nevertheless, even capturing transitions in frailty phenotype or frailty index alone can provide good predictability of mortality risk, which provided good feasibility in adopting the approaches in different clinical settings.

      Acknowledgments

      Prof. Nicole Huang and Prof. Fei-Yuan Hsiao provided critical comments to the study.
      American Journal Experts provided professional editorial services.

      Supplementary Data

      Supplementary Table 1Frailty Index Variables
      Deficit Categories and Component ItemsAnswer (Corresponding Frailty Index Value)
      I: Health status and comorbidities (17 items)
       1. Multimorbidities (confirmed by physician diagnosis): hypertension, diabetes mellitus, heart disease, stroke, cancer, bronchitis or emphysema or asthma, arthritis or rheumatism, peptic ulcer or gastric diseases, hepatobiliary diseases, hip fracture, cataract, renal disease (including stones), gout.≥2 morbidities (1)
      ≤1 morbidity (0)
       2. Hypertension8
      ’8’ denotes that a responder did not have the comorbidity (according to question in the original questionnaire), meaning that the question about daily life impact was not applicable.
      : No (0), 0: Yes, no impact to daily life (0.33),
      1: Yes, some impact on daily life (0.66),
      2: Yes, much impact on daily life (1)
       3. Diabetes mellitus8: No (0), 0: Yes, no impact on daily life (0.33),
      1: Yes, some impact on daily life (0.66),
      2: Yes, much impact on daily life (1)
       4. Heart disease8: No (0), 0: Yes, no impact on daily life (0.33),
      1: Yes, some impact on daily life (0.66),
      2: Yes, much impact on daily life (1)
       5. Stroke8: No (0), 0: Yes, no impact on daily life (0.33),
      1: Yes, some impact on daily life (0.66),
      2: Yes, much impact on daily life (1)
       6. Cancer8: No (0), 0: Yes, no impact on daily life (0.33),
      1: Yes, some impact on daily life (0.66),
      2: Yes, much impact on daily life (1)
       7. Bronchitis or emphysema or asthma8: No (0), 0: Yes, no impact on daily life (0.33),
      1: Yes, some impact on daily life (0.66),
      2: Yes, much impact on daily life (1)
       8. Arthritis or rheumatism8: No (0), 0: Yes, no impact on daily life (0.33),
      1: Yes, some impact on daily life (0.66),
      2: Yes, much impact on daily life (1)
       9. Peptic ulcer or gastric diseases8: No (0), 0: Yes, no impact on daily life (0.33),
      1: Yes, some impact on daily life (0.66),
      2: Yes, much impact on daily life (1)
       10. Hepatobiliary diseases8: No (0), 0: Yes, no impact on daily life (0.33),
      1: Yes, some impact on daily life (0.66),
      2: Yes, much impact on daily life (1)
       11. Hip fracture8: No (0), 0: Yes, no impact on daily life (0.33),
      1: Yes, some impact on daily life (0.66),
      2: Yes, much impact on daily life (1)
       12. Cataract8: No (0), 0: Yes, no impact on daily life (0.33),
      1: Yes, some impact on daily life (0.66),
      2: Yes, much impact on daily life (1)
       13. Renal disease (including stones)8: No (0), 0: Yes, no impact on daily life (0.33),
      1: Yes, some impact on daily life (0.66),
      2: Yes, much impact on daily life (1)
       14. Gout8: No (0), 0: Yes, no impact on daily life (0.33),
      1: Yes, some impact on daily life (0.66)
      2: Yes, much impact on daily life (1)
       15. Self-rated health1: Very poor (1), 2: Poor (0.75), 3: Fair (0.5),
      4: Good (0.25), 5: Very good (0)
       16. Pain0: None (0), 1: Mild (0.25), 2: Moderate (0.5),
      3: Severe (tolerable) (0.75), 4: Very severe (intolerable) (1)
       17. Health status evaluated by spouse1: Very good (0), 2: Good (0.25), 3: Fair (0.5),
      4: Poor (0.75), 5: Very poor (1)
      II: Physical activity, ADL and IADL (22 items)
       1. Fall or accidental injury in past year0: No (0), 1: Yes (1)
       2. Stand for 15 minutes0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       3. Stand for 2 hours0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       4. Squatting0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       5. Raising both hands over head0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       6. Grasping objects with fingers0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       7. Lifting 11–12 kilograms0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       8. Run for 20–30 minutes0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       9. Walking 200–300 m0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       10. Climb 2–3 flights of stairs0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       11. Taking Bath0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       12. Dressing0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       13. Eating0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       14. Get up from bed; stand; sit on the chair0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       15. Moving around the house0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       16. Toileting0: No difficulty (0), 1: Some difficulty, (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       17. Buying personal item0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       18. Managing money0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       19. Riding bus/train on one’s own0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       20. Doing light tasks at home0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       21. Doing physical work at home0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
       22. Making phone calls0: No difficulty (0), 1: Some difficulty (0.33),
      2: Much difficulty (0.66), 3: Can’t do it at all (1)
      III: Cognitive domain (10 items)
       1. Orientation to time (year)0: Wrong (1), 1: Right (0)
       2. Orientation to time (month)0: Wrong (1), 1: Right (0)
       3. Orientation to time (day)0: Wrong (1), 1: Right (0)
       4. Orientation to time (day of the week)0: Wrong (1), 1: Right (0)
       5. Recall home address0: Wrong (1), 1: Right (0)
       6. Recall mother’s family name0: Wrong (1), 1: Right (0)
       7. Recall current president’s name0: Wrong (1), 1: Right (0)
       8. Recall former president’s name0: Wrong (1), 1: Right (0)
       9. Know his/her own age0: Wrong (1), 1: Right (0)
       10. Immediate recall (10 items)Participants answering X items, score = (1- (X/10))
      IV: Psychological (10 items)
       1. Poor appetite0: No (0), 1: Seldom (1 day/week) (0.33),
      2: Sometimes (2–3 days/week) (0.66),
      3: Usually or often (more than 4 days/week) (1)
       2. Everything was an effort0: No (0), 1: Seldom (1 day/week) (0.33),
      2: Sometimes (2–3 days/week) (0.66),
      3: Usually or often (more than 4 days/week) (1)
       3. Could not get going0: No (0), 1: Seldom (1 day/week) (0.33),
      2: Sometimes (2–3 days/week) (0.66),
      3: Usually or often (more than 4 days/week) (1)
       4. Insomnia0: No (0), 1: Seldom (1 day/week) (0.33),
      2: Sometimes (2–3 days/week) (0.66),
      3: Usually or often (more than 4 days/week) (1)
       5. Feels lonely0: No (0), 1: Seldom (1 day/week) (0.33),
      2: Sometimes (2–3 days/week) (0.66),
      3: Usually or often (more than 4 days/week) (1)
       6. Low mood0: No (0), 1: Seldom (1 day/week) (0.33),
      2: Sometimes (2–3 days/week) (0.66),
      3: Usually or often (more than 4 days/week) (1)
       7. Feel others are unfriendly0: No (0), 1: Seldom (1 day/week) (0.33),
      2: Sometimes (2–3 days/week) (0.66),
      3: Usually or often (more than 4 days/week) (1)
       8. Feels happy0: Usually or often (more than 4 days/week) (0),
      1: Sometimes (2–3 days/week) (0.33),
      2: Seldom (1 day/week) (0.66), 3: Never (1)
       9. Life goes well0: Usually or often (more than 4 days/week) (0),
      1: Sometimes (2–3 days/week) (0.33),
      2: Seldom (1 day/week) (0.66), 3: Never (1)
       10. Feels sad0: No (0), 1: Seldom (1 day/week) (0.33),
      2: Sometimes (2–3 days/week) (0.66),
      3: Usually or often (more than 4 days/week) (1)
      V: Life satisfaction, stress, and financial status (10 items)
       1. Satisfaction with life0: No (1), 1: Yes (0)
       2. Satisfaction with what you’re doing0: No (1), 1: Yes (0)
       3. Living in a safe and secure environment0: No (1), 1: Yes (0)
       4. Satisfaction with current financial status1: Very good (0), 2: Good (0.25), 3: Fair (0.5),
      4: Poor (0.75), 5: Very poor (1)
       5. Meeting living expenses1: No difficulty (0), 2: Fair (0.33),
      3: Some difficulty (0.66), 4: Much difficulty (1)
       6. Stress on one’s own finances0: None (0), 1: Sometimes (0.5), 2: Often (1)
       7. Stress on one’s job0: None (0), 1: Sometimes (0.5), 2: Often (1)
       8. Stress on family member’s health, finances, or job0: None (0), 1: Sometimes (0.5), 2: Often (1)
       9. Stress on family’s relationship0: None (0), 1: Sometimes (0.5), 2: Often (1)
       10: Concerns from family and friends1: Very good (0), 2: Good (0.25), 3: Fair (0.5),
      4: Poor (0.75), 5: Very poor (1)
      VI: Sensory (3 items)
       1. Visual impairment0: Can see without glasses (0),
      1: Can see with glasses (0.5),
      2: Visual impairment (1)
       2. Hearing impairment0: Can hear without hearing aids (0),
      1: Can hear with hearing aids (0.5),
      2: Hearing impairment (1)
       3. Oral intake difficulty0: Can eat without denture (0),
      1: Can eat with denture (0.5),
      2: Poor oral intake (1)
      ADL. activities of daily living; IADL, instrumental activities of daily living.
      8’ denotes that a responder did not have the comorbidity (according to question in the original questionnaire), meaning that the question about daily life impact was not applicable.
      Supplementary Table 2ICD-9 Codes Used in Cause of Death Classification
      Cause of Death CategoryICD-9 Codes
      Infectious diseases001–139, 460–466, 473, 480–488, 597, 599
      Malignancy140–239
      Cardiometabolic and cerebrovascular diseases250, 390–417, 430–459
      Organ failure (heart, lung, digestive system, kidney)420–429, 490–496, 500–508, 511–519, 530–579, 580–588
      Others240–246, 251–259, 260–269, 270–279, 280–289, 290–389, 680–709, 710—739, 780–799, 800–999
      Supplementary Table 3Participant Characteristics in 1999 and 2003 (N = 3529)
      Characteristics1999 Wave2003 Wave
      Baseline characteristics
       Age, y67.6 ± 8.671.6 ± 8.6
       Male sex1817 (51.5)1817 (51.5)
       Education, y5.1 ± 4.65.1 ± 4.6
       Marital status
      Married/cohabiting2503 (70.9)2277 (64.5)
       Alcohol consumption
      More than once per week.
      400 (11.4)338 (9.6)
       Current smoker852 (24.2)675 (19.1)
       Medical history
      Self-report with previous physician diagnosis.
      Number of comorbidities1.6 ± 1.62.1 ± 1.8
      Hypertension1148 (32.5)1423 (40.3)
      Diabetes452 (12.8)613 (17.4)
      Heart disease597 (16.9)790 (22.4)
      Stroke149 (4.2)280 (7.9)
      Cancer86 (2.4)121 (3.4)
      Chronic lung disease388 (11.0)501 (14.2)
      Peptic ulcer disease672 (19.0)747 (21.2)
      Hepatobiliary disease256 (7.3)326 (9.2)
      Chronic kidney disease262 (7.4)351 (9.9)
      Frailty assessment
       Frailty phenotype
      Slowness582 (16.5)939 (26.6)
      Weakness212 (6.0)378 (10.7)
      Low physical activity696 (20.3)851 (24.1)
      Exhaustion687 (20.3)700 (21.6)
      Weight loss178 (5.2)194 (5.9)
       Robustness2041 (57.8)1791 (50.8)
       Prefrailty1263 (35.8)1349 (38.2)
       Frailty225 (6.4)389 (11.0)
       Summed frailty phenotype score (0–5)0.7 ± 0.90.9 ± 1.1
       Frailty index0.15 ± 0.110.19 ± 0.15
      Data are shown as mean ± SD or number (percentage).
      More than once per week.
      Self-report with previous physician diagnosis.
      Supplementary Table 4Frailty Index Percentage Change and Cause-specific Mortality
      Cause of DeathNumber of Deaths (%)Model 1Model 2
      Infection
       Total72/3529 (2.0)
       Decreased frailty index10/1304 (0.8)Ref
       Moderately increased frailty index (<50%)27/997 (2.7)4.1 (1.8–9.4)<.001
       Rapidly increased frailty index (≥50%)35/1228 (2.9)5.4 (2.4–12.6)<.001
      Malignancy
       Total168/3529 (4.8)
       Decreased frailty index44/1304 (3.4)RefRef
       Moderately increased frailty index (<50%)40/997 (4.0)1.0 (0.6–1.6).9721.0 (0.7–1.6)0.846
       Rapidly increased frailty index (≥50%)84/1228 (6.8)1.8 (1.2–2.7).0091.9 (1.2–2.9).006
      Cardiometabolic/cerebrovascular
       Total189/3529 (5.4)
       Decreased frailty index51/1304 (3.9)RefRef
       Moderately increased frailty index (<50%)53/997 (5.3)1.2 (0.8–1.8).4741.1 (0.7–1.7).633
       Rapidly increased frailty index (≥50%)85/1228 (6.9)2.7 (1.8–4.1)<.0012.2 (1.5–3.4)<.001
      Organ failure
       Total130/3529 (3.7)
       Decreased frailty index36/1304 (2.8)RefRef
       Moderately increased frailty index (<50%)32/997 (3.2)1.0 (0.6–1.6).9141.0 (0.6–1.6).930
       Rapidly increased frailty index (≥50%)62/1228 (5.0)1.4 (0.8–2.4).1981.5 (0.8–2.5).176
      Others
       Total96/3529 (2.7)
       Decreased frailty index27/1304 (2.1)Ref
       Moderately increased frailty index (<50%)23/997 (2.3)1.1 (0.6–2.0).744
       Rapidly increased frailty index (≥50%)46/1228 (3.7)2.0 (1.2–3.6).012
      Model 1:Adjust age, gender, education, smoke, alcohol, baseline frailty index, total comorbidity number.
      Model 2: Malignancy: model 1 + baseline malignancy.
      Cardiometabolic: model 1 + baseline DM, HTN, heart disease, stroke.
      Organ failure: model 1+ baseline heart disease, lung disease, hepatobiliary disease, kidney disease.
      Supplementary Table 5Combination of Frailty Phenotype and Frailty Index Change (by Percentage) and All-Cause Mortality
      Decreased Frailty Index (n = 168/1304)Moderately Increased Frailty Index (<50%)

      (n = 175/997)
      Rapidly Increased Frailty Index (≥50%)

      (n = 312/1228)
      n (%)OR (95% CI)P ValuenOR (95% CI)P ValuenOR (95% CI)P Value
      Maintained robustness or prefrailty68/704 (9.7)156/492 (11.4)173/533 (13.7)1
      Worsened frailty28/194 (14.4)1.1 (0.7–1.7).79759/296 (19.9)1.4 (1.0–2.1).076209/575 (36.3)2.7 (2.0–3.6)<.001
      Improved frailty60/383 (15.7)0.7 (0.4–1.2).21243/180 (23.9)1.1 (0.7–1.8).73421/104 (20.2)0.8 (0.5–1.5).497
      Maintained frailty12/23 (52.2)1.1 (0.5–2.5).88517/29 (58.6)1.8 (0.9–3.8).1109/16 (56.3)1.3 (0.5–3.1).560
      Model: Adjust age, sex, education, marital status, alcohol, smoke, total comorbidity numbers, baseline frailty phenotype.
      Supplementary Table 6Combination of Frailty Phenotype and Frailty Index Change (by Percentage) and All-Cause Mortality
      Maintained Robustness or Prefrailty (n = 197/1729)Worsened Frailty (n = 296/1065)Improved Frailty (n = 124/667)Maintained Frailty (n = 38/68)
      n (%)OR (95% CI)P Valuen (%)OR (95% CI)P Valuen (%)OR (95% CI)P Valuen (%)OR (95% CI)P Value
      Decreased frailty index68/704 (9.7)128/194 (14.4)160/383 (15.7)112/23 (52.2)1
      Moderately increased frailty index56/492 (11.4)1.0 (0.7–1.5).82559/296 (19.9)1.5 (0.9–2.3).09343/180 (23.9)1.8 (1.2–2.7).00717/29 (58.6)1.8 (0.8–4.1).139
      Rapidly increased frailty index73/533 (13.7)1.3 (0.9–1.9).160209/575 (36.3)4.3 (2.8–6.7)<.00121/104 (20.2)2.0 (1.1–3.4).0179/16 (56.3)1.7 (0.5–5.1).366
      Model: Adjust age, sex, education, marital status, alcohol, smoke, total comorbidity numbers, baseline frailty index.
      Supplementary Table 7The Impact of Baseline Frailty Phenotype and Frailty Phenotype Transition on All-Cause Mortality
      HR (95% CI)P Value
      Frailty phenotype transition
       Maintained robustness or prefrailtyRef
       Worsened frailty1.9 (1.4–2.4)<.001
       Improved frailty0.9 (0.7–1.3).774
       Maintained frailty1.4 (0.8–2.4).282
      Baseline frailty phenotype
       RobustRef
       Prefrail1.5 (1.1–2.0).021
       Frail3.0 (1.8–5.0)<.001
      Baseline frailty phenotype∗ frailty phenotype transition
       Baseline frailty phenotype∗ frailty phenotype transitionRef
       Baseline frailty phenotype (1)∗ frailty phenotype transition1.2 (0.8–1.8).285
       Baseline frailty phenotype (2)∗ frailty phenotype transition0.7 (0.3–1.8).455
      Model: Include age, sex, education, marital status, alcohol, smoke, number of comorbidities, baseline frailty phenotype, frailty phenotype transition, and interaction term (baseline frailty phenotype) ∗(frailty phenotype transition).

      References

        • Clegg A.
        • Young J.
        • Iliffe S.
        • et al.
        Frailty in elderly people.
        Lancet. 2013; 381: 752-762
        • Draznin B.
        • Aroda V.R.
        • et al.
        • American Diabetes Association Professional Practice Committee
        13. Older Adults: Standards of Medical Care in Diabetes-2022.
        Diabetes Care. 2022; 45: S195-S207
        • Williams B.
        • Mancia G.
        • Spiering W.
        • et al.
        2018 ESC/ESH Guidelines for the management of arterial hypertension.
        Eur Heart J. 2018; 39: 3021-3104
        • Kojima G.
        Frailty as a predictor of future falls among community-dwelling older people: a systematic review and meta-analysis.
        J Am Med Dir Assoc. 2015; 16: 1027-1033
        • Kojima G.
        • Iliffe S.
        • Morris R.W.
        • et al.
        Frailty predicts trajectories of quality of life over time among British community-dwelling older people.
        Qual Life Res. 2016; 25: 1743-1750
        • Rockwood K.
        • Mitnitski A.
        • Song X.
        • et al.
        Long-term risks of death and institutionalization of elderly people in relation to deficit accumulation at age 70.
        J Am Geriatr Soc. 2006; 54: 975-979
        • Fried L.P.
        • Tangen C.M.
        • Walston J.
        • et al.
        Frailty in older adults: evidence for a phenotype.
        J Gerontol A Biol Sci Med Sci. 2001; 56: M146-M156
        • Chang S.F.
        • Lin P.L.
        Frail phenotype and mortality prediction: a systematic review and meta-analysis of prospective cohort studies.
        Int J Nurs Stud. 2015; 52: 1362-1374
        • Kojima G.
        • Iliffe S.
        • Walters K.
        Frailty index as a predictor of mortality: a systematic review and meta-analysis.
        Age Ageing. 2018; 47: 193-200
        • Thompson M.Q.
        • Theou O.
        • Tucker G.R.
        • et al.
        Recurrent measurement of frailty is important for mortality prediction: findings from the North West Adelaide Health Study.
        J Am Geriatr Soc. 2019; 67: 2311-2317
        • Kulmala J.
        • Nykanen I.
        • Hartikainen S.
        Frailty as a predictor of all-cause mortality in older men and women.
        Geriatr Gerontol Int. 2014; 14: 899-905
        • Wang M.C.
        • Li T.C.
        • Li C.I.
        • et al.
        Frailty, transition in frailty status and all-cause mortality in older adults of a Taichung community-based population.
        BMC Geriatr. 2019; 19: 26
        • Stolz E.
        • Hoogendijk E.O.
        • Mayerl H.
        • et al.
        Frailty changes predict mortality in four longitudinal studies of aging.
        J Gerontol A Biol Sci Med Sci. 2021; 76: 1619-1626
        • Shi S.M.
        • Olivieri-Mui B.
        • McCarthy E.P.
        • et al.
        Changes in a frailty index and association with mortality.
        J Am Geriatr Soc. 2021; 69: 1057-1062
        • Rockwood K.
        • Andrew M.
        • Mitnitski A.
        A comparison of two approaches to measuring frailty in elderly people.
        J Gerontol A Biol Sci Med Sci. 2007; 62: 738-743
        • Lohman M.C.
        • Sonnega A.J.
        • Resciniti N.V.
        • et al.
        Frailty phenotype and cause-specific mortality in the United States.
        J Gerontol A Biol Sci Med Sci. 2020; 75: 1935-1942
        • Grabovac I.
        • Haider S.
        • Mogg C.
        • et al.
        Frailty status predicts all-cause and cause-specific mortality in community dwelling older adults.
        J Am Med Dir Assoc. 2019; 20: 1230-1235.e1232
        • Fan J.
        • Yu C.
        • Guo Y.
        • et al.
        Frailty index and all-cause and cause-specific mortality in Chinese adults: a prospective cohort study.
        Lancet Public Health. 2020; 5: e650-e660
      1. Health Promotion Administration, Ministry of Health and Welfare, Taiwan, Taiwan Longitudinal Study on Aging (TLSA).
        2015
        • Hwang A.C.
        • Lee W.J.
        • Huang N.
        • et al.
        Longitudinal changes of frailty in 8 years: comparisons between physical frailty and frailty index.
        BMC Geriatr. 2021; 21: 726
        • Chen C.Y.
        • Wu S.C.
        • Chen L.J.
        • et al.
        The prevalence of subjective frailty and factors associated with frailty in Taiwan.
        Arch Gerontol Geriatr. 2010; 50: S43-S47
        • Searle S.D.
        • Mitnitski A.
        • Gahbauer E.A.
        • et al.
        A standard procedure for creating a frailty index.
        BMC Geriatr. 2008; 8: 24
        • Jang I.Y.
        • Jung H.W.
        • Lee H.Y.
        • et al.
        Evaluation of clinically meaningful changes in measures of frailty.
        J Gerontol A Biol Sci Med Sci. 2020; 75: 1143-1147
        • Li C.Y.
        • Al Snih S.
        • Karmarkar A.
        • et al.
        Early frailty transition predicts 15-year mortality among nondisabled older Mexican Americans.
        Ann Epidemiol. 2018; 28: 362-367.e363
        • Xue Q.L.
        • Bandeen-Roche K.
        • Tian J.
        • et al.
        Progression of physical frailty and the risk of all-cause mortality: is there a point of no return?.
        J Am Geriatr Soc. 2021; 69: 908-915
        • Li X.
        • Ploner A.
        • Karlsson I.K.
        • et al.
        The frailty index is a predictor of cause-specific mortality independent of familial effects from midlife onwards: a large cohort study.
        BMC Med. 2019; 17: 94
        • Cespedes Feliciano E.M.
        • Hohensee C.
        • Rosko A.E.
        • et al.
        Association of prediagnostic frailty, change in frailty status, and mortality after cancer diagnosis in the Women's Health Initiative.
        JAMA Netw Open. 2020; 3: e2016747
        • Lynn J.
        Perspectives on care at the close of life. Serving patients who may die soon and their families: the role of hospice and other services.
        JAMA. 2001; 285: 925-932
        • Yang X.
        • Lupon J.
        • Vidan M.T.
        • et al.
        Impact of frailty on mortality and hospitalization in chronic heart failure: a systematic review and meta-analysis.
        J Am Heart Assoc. 2018; 7: e008251
        • Puhan M.A.
        • Siebeling L.
        • Zoller M.
        • et al.
        Simple functional performance tests and mortality in COPD.
        Eur Respir J. 2013; 42: 956-963
        • Bao Y.
        • Dalrymple L.
        • Chertow G.M.
        • et al.
        Frailty, dialysis initiation, and mortality in end-stage renal disease.
        Arch Intern Med. 2012; 172: 1071-1077
        • Fried L.P.
        • Cohen A.A.
        • Xue Q.L.
        • et al.
        The physical frailty syndrome as a transition from homeostatic symphony to cacophony.
        Nat Aging. 2021; 1: 36-46
        • Cesari M.
        • Gambassi G.
        • van Kan G.A.
        • et al.
        The frailty phenotype and the frailty index: different instruments for different purposes.
        Age Ageing. 2014; 43: 10-12
        • Rockwood K.
        • Song X.
        • MacKnight C.
        • et al.
        A global clinical measure of fitness and frailty in elderly people.
        CMAJ. 2005; 173: 489-495
        • Theou O.
        • Brothers T.D.
        • Mitnitski A.
        • et al.
        Operationalization of frailty using eight commonly used scales and comparison of their ability to predict all-cause mortality.
        J Am Geriatr Soc. 2013; 61: 1537-1551
        • Woo J.
        • Leung J.
        • Morley J.E.
        Comparison of frailty indicators based on clinical phenotype and the multiple deficit approach in predicting mortality and physical limitation.
        J Am Geriatr Soc. 2012; 60: 1478-1486