If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
We aimed to explore predictors of sustained transitions (those that are maintained for an extra follow-up) between robustness and prefrailty in both directions.
Design
Longitudinal population-based cohort.
Setting and Participants
Community-dwelling Spaniards 65 years or older from the Toledo Study of Healthy Ageing.
Methods
The Fried's frailty phenotype was measured over 3 waves (2006–2009, 2011–2013, and 2014–2017). Multiple logistic regressions compared individuals following the pattern robust-prefrail-prefrail with those who remained robust across waves, and those following the pattern prefrail-robust-robust with those who remained prefrail, for sociodemographic, clinical, life-habits, dependency for activities of daily living, upper and lower extremities’ strength variables. The Fried's items of those who remained prefrail and those who became robust were compared.
Results
Mean age was 72.3 years (95% CI: 71.8–72.8) and 57.9% (52.7%–63.0%) were women. After multivariate adjustment, predictors (apart from age) of the sustained transition robustness-prefrailty were as follows: number of drugs taken (odds ratio: 1.37; 95% CI: 1.14–1.65), not declaring the amount of alcohol consumed (8.32; 1.78–38.88), and grip strength (0.92 per kg; 0.86–0.99). Predictors of the sustained transition prefrailty-robustness were as follows: drinking alcohol (0.2; 0.05–0.83), uricemia (0.67; 0.49–0.93), number of chair stands in 30 seconds (1.14; 1.01–1.28), and grip strength (1.12; 1.05–1.2). Low grip strength was associated with a lower probability of regaining robustness.
Conclusions and Implications
Prediction of sustained transitions between the first stages of frailty development can be achieved with a reduced number of variables and noting whether the Fried's item leading to a diagnosis of prefrailty is low grip strength. Our results suggest the need to intensify interventions on deprescription, quitting alcohol, and strengthening of upper and lower limbs.
have analyzed transitions in the frailty phenotype states between 2 waves. Indeed, the papers that have revised more than 2 waves have limited themselves to analyzing pairs of consecutive measurements of frailty status
Although relevant, these changes can be due to measurement error or be just transient, having in mind the dynamic nature of frailty in which up to 40% of the people experience some sort of transitions, including transitions to less frail states.
This usual way of assessing transitions has the risk of overdetection. Consider for example the difference between an individual who transitions across several waves from robustness to prefrailty and remains prefrail, and one who first transitions to prefrailty and then back to robustness again. In a 2-waves analysis, both first transitions would be included in the same category, although only the former one would have experienced a true transition.
we analyzed many possible predictors of changes in the frailty status between the basal and the first follow-up of a cohort of community-dwelling older people. The addition of a now available second follow-up allows us to determine whether changes were maintained over time. Predictors of these sustained changes may not be necessarily the same as those of more transient ones and, because of its stability, they are more relevant for research and clinical care purposes. The same reasoning applies to the study of the prognostic capacity of the different frailty phenotype items.
We focus this study on change in the initial stages of frailty development in its 2 potential directions: transitioning from robustness to prefrailty and from prefrailty to robustness. Knowing the determinants of these changes is especially relevant not only because of their frequency, but because these stages of the transitions are the most amenable to successful interventions.
The aim of this paper was to study the predictors of maintained transitions, that is, changes between 2 waves that persisted in the next one, between robustness and prefrailty in both directions throughout 8 years in a cohort of community-dwelling Spaniards. In addition, we wanted to know if any of the frailty phenotype items that led to a diagnosis of prefrailty was associated with a better or worse evolution of prefrailty.
Methods
Data for this paper come from the Toledo Study on Healthy Ageing (TSHA), a population-based, prospective longitudinal cohort study of individuals aged >65 years living in the province of Toledo (Spain).
Individuals living in nursing homes were excluded. Data belong to the first (2006–2009), second (2011–2013), and third (2014–2017) waves of the study. In each wave, all attempts were made to interview individuals in the same order they were interviewed in the previous one. Frailty was measured according to the criteria of Fried et al.
Weakness was defined as the worse quintile of maximum grip strength on the dominant hand adjusted for sex and body mass index (BMI), and slowness as the worse quintile of gait speed, adjusted for sex and height. Individuals were considered to have “low energy” when they provided a positive answer to any of the following 2 questions from the Center for Epidemiologic Studies Depression Scale (CES-D scale)
: “I felt that anything I did was a big effort” and “I felt that I could not keep on doing things” at least 3 to 4 days a week. Low physical activity was operationalized as being in the worse quintile of the Physical Activity Scale for the Elderly (PASE) score
and weight loss as having experienced unintentional weight loss of 4.5 kg or more in the past year. Quintiles were calculated for the baseline wave and the cutoff values found used in the second and third waves. According to the number of criteria present, individuals were classified as nonfrail (0 criteria), prefrail (1–2 criteria), and frail (3–5 criteria). We checked all possible combinations of the frailty phenotype and death throughout the 3 waves but considered only changes between frailty states that were maintained over more than 1 wave. In this paper, we report on the 2 most frequent sustained transitions: prefrail-robust-robust and robust-prefrail-prefrail.
Predictors were measured at baseline. (1) Self-reported predictors were age, gender, education level, civil status, and living arrangements. (2) A count of the number of the following diseases reported in medical records or by the participant: myocardial infarction–cardiac arrest, heart failure, angor, atrial fibrillation, peripheral vascular disease, hypertension, stroke, diabetes mellitus, Parkinson disease, chronic obstructive pulmonary disease, asthma, peptic ulcer disease, renal failure–dialysis–renal transplant, rheumatoid arthritis, osteoarthritis of hips or knees, cirrhosis–severe hepatic disease, and cancer (including blood and skin cancers). (3) Self-perceived health, categorized as good or not good. (4) Number of depressive symptoms according to the Geriatric Depression Scale.
(5) Alcohol (categorical, where those drinking at risk were men who drank 28 or more units of alcohol per week and women who drank 17 or more units per week) and tobacco consumption. (6) Physical activity according to the PASE scale.
(7) Number of falls in the previous year and need of any medical assistance for the fall. (8) Visual impairment that made it difficult to read or sew, carry out activities of daily living, or almost produced blindness, and hearing impairment that prevented from taking part in a group conversation, an individual conversation at a normal volume and distance, or maintaining a conversation altogether. (9) Dependence for any basic (BADL-Katz index)
(2) BMI, obtained through dividing the weight in kilograms (measured with a SECA precision scale) by the height in meters squared (measured with a stadiometer on a wall without skirting board) and categorized as underweight (BMI <18.5), normal weight (≥18.5<25), overweight (≥25<30), and obesity degrees I (≥30<35), II (≥35<40), and III (≥40). (3) A count of the drugs presented by individuals as currently consumed and with an ATC code. (4) Gait speed (meters/second), measured along a 3-meter distance according to the standards of the Short Physical Performance Battery.
A short physical performance battery assessing lower extremity function: association with self-reported disability and prediction of mortality and nursing home admission.
(5) Grip strength (kilograms), measured as the best out of 3 trials with a Jamar hydraulic dynamometer in the dominant hand, according to the standards of the Hispanic Established Populations for Epidemiologic Studies of the Elderly.
(6) Number of stands from a chair in 30 seconds with arms folded over the chest. (7) The best of 3 trials of shoulder (abduction), hip (flexion), and knee (extension) strength measured with a Muscular Mass Test (MMT) device. (8) Hemoglobin (g/dL) and anemia (having less than 13 g/dL in men or 12 g/dL in women), fasting glycemia (mg/dL) and hyperglycemia (more than 126 mg/dL), uricemia (mg/dL), and glomerular filtration (mL/min per 1.73 m2), calculated from serum creatinine, age, and sex according to the equation in the paper by Levey et al.
respectively. No imputation was performed for the frailty variable. The predictors with missing values were imputed, using as predictors the rest of the variables, transformed if needed. Continuous variables were imputed with predictive mean matching and categorical variables with a multinomial logit model, which ensure that imputed values are within original ranges. The number of imputed samples was 42. Variables were described through means or percentages with their 95% CIs depending on the type of variable. Because of the complex survey structure of the data, standard test (means comparison, chi square) for the comparisons between types of trajectories could not be performed and were substituted by the coefficient of the independent variable “type of trajectory” in simple linear (quantitative predictors) or logistic (binary predictors) regressions.
Two multivariate logistic regressions were performed. In the first analysis, individuals that remained robust at the 3 measurement points were compared with those individuals who became prefrail in the second wave and remained so in the third one. In the second analysis, those who remained prefrail at the 3 waves were compared with those who became robust at the second wave and remained so at the third.
For both analyses, 4 models were calculated: First, a bivariate model; second, the same model adding the variable age; third, a model with all variables that showed a P value less than .2 in the previous model; and finally, a reduced model in which we removed sequentially nonsignificant variables in the full model whose deletion did not change by more than 10% the coefficients of significant (P < .05) or close to the significance level variables. We studied linearity with scatterplot smoothing and density estimation, and tried splines and power, logarithmic, and inverse transformations when needed.
Finally, we studied which of the items of Fried et al
were associated with the persistent recovery of robustness compared with a persistent prefrail state, first with a χ2 test and then with a logistic regression of the different presenting items of Fried et al
A total of 2475 individuals were eligible. The frailty phenotype could be measured at baseline in 1750 and complete information on the 3 waves was obtained from 949. The most frequent trajectories experienced by these individuals are presented in Figure 1. The figure shows that transitions back to the previous state, which would be indistinguishable from more persistent transitions in a 2-waves analysis, are very frequent: 58 individuals experienced the pattern robust-prefrail-prefrail, compared with 57 following the pattern robust-prefrail-robust; 80 individuals followed the pattern prefrail-robust-robust, and 50 experienced the prefrail-robust-prefrail pattern.
Fig. 1Flow chart of the follow-up assessment of the Toledo Study on Healthy Ageing and most frequent transitions. The abbreviations represent the sequence of frailty states at each wave of the study. R = Robust, P = Prefrail, F = Frail.
Compared with those who could not be assessed for frailty at baseline or whose follow-up information was incomplete, those with full follow-up information (but who were not dead) were 2 years younger (P < .001), more frequently married but less frequently widowed (73.3 vs 67.8%, 19% vs 24.9%, P = .044), less dependent for BADL (6.2% vs 18.4%, P < .001), and with less frequent visual impairment (14.2% vs 17.7%, P = .013). There were no differences in gender, education level, number of drugs taken, self-perceived health, or hearing impairment.
Variables with the highest proportion of missing values (Table 1) were those that required a blood extraction (12%), followed by cognitive status (10%), depression (10%), and number of diseases (8%). Descriptive values were extremely similar for the nonimputed (data not shown) and imputed samples (Table 1). Mean age was 72.3 years (95% CI: 71.8–72.8) and 57.9% (52.7–63) were women. The education level was very low and most individuals were married. Almost three-quarters referred a good level of health, the average number of diseases was 1.6 (1.3–1.8) and of drugs 3.9 (3.7–4.2). The mean BMI was in the overweight range, with only 13.4% (10.1–16.7) having a normal weight. The great majority had never consumed alcohol or tobacco. Vision or hearing impairment affected 12%, and almost half were dependent for at least 1 IADL, although only 3.7% (1.8%–5.5%) were dependent for at least 1 BADL. The average gait speed was 2 decimeters below the considered normal speed. Average biochemic levels were in the normal range.
Table 1Description of Variables and Proportion of Missing Values at Baseline (Wave 2006–2009)
% Missing (n = 387)
Imputed Sample (n = 387)
Age
0
72.3 (71.8–72.8)
Women
0
57.9 (52.7–63)
Education
0.5
No studies
62.3 (52.1–72.5)
Less than primary
17.9 (12.2–23.7)
Primary
9.3 (5.2–13.4)
Secondary
4.9 (2–7.8)
University
5.5 (2.1–8.8)
Civil status
0
Married
74.9 (70.5–79.3)
Widow
17.3 (13.9–20.7)
Single
6.2 (3.8–8.6)
Divorced/separated
1.6 (0–3.1)
Living arrangements
0.2
With spouse
73.4 (68.6–78.1)
Couple living with son/daughter
1.8 (0.4–3.2)
With son
4.1 (2.5–5.8)
With daughter
2.3 (1–3.7)
With brother/nephew
1.6 (0.3–2.8)
Alone
16.8 (13.1–20.5)
Number of diseases
7.5
1.6 (1.3–1.8)
Number of drugs
0
3.9 (3.7–4.2)
Self-perceived health
0
Good
73.9 (68.3–79.5)
Not good
26.1 (20.5–31.7)
MMSE score
10.1
25 (24.4–25.6)
GDS score
9.8
2 (1.6–2.3)
Body Mass Index (kg/height in m2)
0
29.5 (28.9–30)
Body Mass Index
0
Normal weight (≥18.5,<25)
13.4 (10.1–16.7)
Overweight (≥25, <30)
45.2 (41–49.4)
Obesity I (≥30, <35)
32.3 (27–37.6)
Obesity II (≥35,<40)
6.7 (4.4–9.1)
Obesity III (≥40)
2.3 (1–3.8)
Alcohol
2.6
Teetotaller
68.9 (63.1–74.8)
Former drinker
8.2 (5.2–11.2)
Not at risk
10.6 (4.8–16.4)
At risk
0.8 (0–2)
Does not declare consumption
11.4 (6.8–16)
Smoking
1
Never smoker
69.2 (63.9–74.5)
Former smoker
22.4 (18.2–26.6)
Current smoker
8.4 (5.6–11.2)
PASE
0
81.7 (74.5–88.9)
Number of falls past year
0.5
0.3 (0.2–0.3)
Falls requiring health care
0.5
No falls
82.8 (78.3–87.3)
Yes
6.5 (4.3–8.6)
No
10.7 (6–15.5)
Vision impairment
0
12.1 (8.5–15.8)
Hearing impairment
0
12.7 (7.7–17.6)
Dependent for BADL
1
3.7 (1.8–5.5)
Dependent for IADL
4.7
48.7 (42–55.3)
Gait speed (m/s)
0
0.63 (0.61–0.66)
Chair stands
1.8
11.1 (10.6–11.6)
Grip strength (kg)
0
24.5 (23.4–25.6)
Shoulder strength (kg)
1
17.2 (16–18.3)
Hip strength (kg)
0
20.4 (18.9–21.9)
Knee strength (kg)
0
14.7 (13.5–16)
Hemoglobin (g/dL)
12.4
14.4 (14.3–14.6)
Glycemia (mg/dL)
12.1
105.2 (101.7–108.8)
Uricemia (mg/dL)
12.1
5 (4.9–5.1)
Glomerular filtration (mL/min/1.73 m2)
12.1
80.8 (78.9–82.8)
Mean/percentage (95% confidence interval).
GDS, Geriatric Depression Scale; MMSE, Mini Mental Status Examination; PASE, Physical Activity Scale for the Elderly.
Table 2 shows the differences between those who remained robust for the full follow-up and those who developed persistent prefrailty. The latter were older, took more drugs, had more depressive symptoms, did not declare more often the amount of alcohol consumed, performed less physical activity, and had a lower grip strength and glomerular filtration. In the reduced (and full) multivariate logistic model (Table 3), only an increasing age over 77 years (odds ratio, 95% CI: 2.1, 1.51–2.91), each extra drug taken (1.37, 1.14–1.65), and not declaring the amount of alcohol consumed (8.32, 1.78–38.88) significantly increased the risk of developing persistent prefrailty, whereas each additional kilogram of grip strength (0.92, 0.86–0.99) protected from experiencing that transition.
Table 2Comparison of Individuals Who Remained Robust and Transitioned to Prefrailty and Remained Prefrail, and of Individuals Who Remained Prefrail and Transitioned to Robustness and Remained Robust, in 3 Waves of the Toledo Study of Healthy Ageing
GDS, Geriatric Depression Scale; MMSE, Mini Mental Status Examination; PASE, Physical Activity Scale for the Elderly; PPP, prefrail over the 3 waves; PRR, prefrail-robust-robust in the 3 waves; RPP, robust-prefrail-prefrail in the 3 waves; RRR, robust over the 3 waves.
∗ The survey package does not allow to obtain the P value for a set of dummy variables.
Table 2 also shows the comparison between those who remained prefrail and those who regained robustness in the 2 latter waves. Those who became robust were younger; had fewer diseases; were less frequently dependent for IADL; performed more chair stands; had more strength in their hands, shoulders, hips, and knees; and had a lower uricemia and higher glomerular filtration. Table 4 shows that, in the fully adjusted model, each additional year of age (0.8, 0.72–0.89), each additional milligram per deciliter of uricemia (0.67, 0.49–0.93), and drinking alcohol (0.2, 0.05–0.83) significantly decreased the chances of becoming robust, whereas each additional chair stand (1.14, 1.01–1.28) and kilogram of grip strength (1.12, 1.05–1.2) increased them.
Table 4Logistic Models of the Persistent Transition From Prefrailty to Robustness
There was a statistical association (P = .002) between recovering or not from prefrailty and the pattern of frailty item(s) in the first wave. Those who remained prefrail along the 3 waves presented more often with a low grip strength (33.8% vs 7.6%, P < .001) and less often with low physical activity (9.2% vs 21.5%, P = .025) than those who recovered. The rest of the differences were not statistically significant: 2 of the items of Fried et al,
including low grip strength 13.8 vs 6.3, P = .182; 2 items excluding low grip strength 7.7% vs 12.7%, P = .340; slow gait speed 18.5 vs 22.8, P = .524; exhaustion 9.2% vs 17.7%, P = .146; weight loss 7.7 vs 11.4, P = .401. After adjusting for age and gender, compared with those with 2 frailty items not including the low grip strength item that was the second most frequent category, those with low grip strength had an odds ratio of 0.66 (95% CI: 0.53–0.82; P < .001) of becoming robust, those with low grip strength and an additional item an odds ratio of 0.73 (95% CI: 0.55–0.98; P = .045), and those with low physical activity an odds ratio of 1.02 (95% CI: 0.76–1.35; P = .914). No other item of Fried et al
To our knowledge, this is the study that, to date, has analyzed more predictive factors of transitions between 2 frailty states that persist over more than 1 follow-up. Predictors are not that many, and several of them are common to those predicting changes between 2 waves
providing consistency to the findings: age, number of drugs taken, not reporting the amount of alcohol consumed, and grip strength for the transition from robustness to prefrailty, and age, grip strength, drinking alcohol, number of chair stands, and uricemia for the reverse pathway. In addition, we established that a diagnosis of prefrailty based on the presence of low grip strength is a predictor of a worse evolution of prefrailty.
Polypharmacy has been reported as a risk factor for the development of frailty even after adjustment for multimorbidity and other variables but has been usually measured as a categorical variable with a cutoff of ≥5
Effects of changes in number of medications and drug burden index exposure on transitions between frailty states and death: the Concord Health and Ageing in Men Project cohort study.
found a hazard ratio of 1.04 (95% CI: 1.00–1.09) for the transition from robustness to prefrailty for each additional drug consumed. It was lower than ours, but their study was carried out only in men and did not study persistent transitions.
Our results point to the need to change the focus when dealing with this preventable frailty risk factor from a dichotomous concept of polypharmacy, which has been defined arbitrarily in at least 111 different numerical ways,
toward the consideration of the actual number of drugs used. Under this light, any reduction in the number of unnecessarily prescribed drugs (ie, deprescription) could have the potential of reducing the risk of developing prefrailty. Supporting this claim, a recent meta-analysis summarized the positive effects on several health outcomes of deprescription
and a study specifically reported an average reduction of 1.35 points (95% CI: −2.22 to −0.48) in the Edmonton Frail Scale score, which has a range of 17 points, after implementing a deprescription intervention in residential aged care facilities.
found that nonhazardous drinking in older age was associated with a significantly lower incidence of prefrailty compared with no alcohol consumption, whereas hazardous consumption showed a nonsignificant lower incidence. Ortolá et al
found that only heavy drinkers showed a decreased risk of developing frailty. Although the 3 studies adjusted for several health conditions, there is room for the results being explained by unaccounted for reverse causality, whereby older adults with health problems choose to limit or reduce their alcohol consumption because of ill health even before the presentation of frailty. A healthy survivor effect, in which only those drinkers who are healthy reach old age, cannot be ruled out either. Strandberg et al
did not find an association of any level of alcohol consumption in older age with prefrailty in a study carried out only in men, whereas high alcohol consumption in midlife was significantly associated with risk of both prefrailty and frailty in old age.
Our study found that drinkers had a significantly lower likelihood of regaining robustness compared with abstainers. This finding should be mainly attributed to moderate drinkers because in the database before imputation, at-risk individuals (men and women who drank at least 28 and 17 units of alcohol per week, respectively) were only 2 among those who remained robust, 1 among those who remained prefrail, and nil in the groups that transitioned. Trevisan et al
found the opposite effect of low-moderate drinking for this transition between 2 waves. In our study, drinking was associated with a higher nonsignificant risk of transitioning to prefrailty and the risk was almost 3 times higher and statistically significant for those not declaring the amount of alcohol consumed. It is hard to say whether the latter ones were moderate or at-risk consumers.
The rest of the results give even more relevance to some of the predictors of change between 2 waves that our group described recently.
Grip strength would be a predictor of persistent changes in both directions between robustness and prefrailty, in addition to predicting transitions between 2 waves. This is not surprising considering that it is an item of the frailty phenotype, but it is remarkable that other items that contribute to the scale, like gait speed and physical activity, did not show any association. Nevertheless, we have confirmed that a higher number of chair stands, a measure of lower limb power, predicts a sustained recovery from prefrailty over a longer period. We have not corroborated our previous finding that the association of the number of chair stands was only evident above 5. This suggests that the 2 commonest ways to measure this type of power, time to perform 5 chair stands and the number of chair stands performed in 30 seconds, may be used to predict frailty transitions over time.
The finding about the relationship between a lower uricemia and recovery from prefrailty is aligned with what was reported some years ago in other Spanish cohort of older people (ENRICA), in which the authors found a direct association between high uric acid levels and risk of developing frailty in a follow-up of 3.5 years.
Risk factors and protective factors associated with incident or increase of frailty among community-dwelling older adults: A systematic review of longitudinal studies.
which would justify the deleterious effects of high levels of uric acid and the protective effects of low levels that increase the likelihood of recovery from prefrailty.
presented by those who are prefrail, but not the number of items, may predict the future development of prefrailty. Presenting with low grip strength, alone or combined with any other item, is associated with a bad prognosis. This is in agreement with the association between a stronger grip and the recovery to robustness commented above. Xue et al
did not find any association for any item, but their results refer to changes between 2 waves only. Low grip strength is a diagnostic criterion of sarcopenia
that may be differentiating prefrail sarcopenic individuals from prefrail nonsarcopenic ones. Our group has already suggested the existence of 2 types of frailty, sarcopenic and nonsarcopenic frailty, with different health outcomes.
The main strength of our analysis is that it studied several predictors, including many functional measures, of sustained transitions over more than 1 follow-up of a well-studied frailty scale in community-dwelling individuals. We have limited our analyses to the first transitions in the frailty continuum, which were the most frequent ones and where interventions are more effective. It would have been interesting to study the predictors of other persistent transitions, but only 12 prefrail individuals and 4 robust individuals transitioned to a persistent frail state, and, among the frail individuals, only 6 transitioned to sustained prefrailty and 1 to a robust state, which precluded carrying out multivariable analyses. Loss to follow-up other than death is our major limitation. This high attrition rate is not surprising considering that this is a cohort mostly of older people and that the total follow-up time was approximately 8 years. Similar rates have been reported in other long-term epidemiological studies of older people.
We have found evidence that those not followed-up were older and with more visual and BADL impairment. Therefore, they probably would have developed the most advanced transitions between prefrailty and frailty and not the ones studied in this paper. Because our objective was not to provide frequencies of the different transitions, but study the variables associated with them, we consider that our results remain generalizable to any population transitioning between robustness and prefrailty. Because of the design of the fieldwork, intervals between assessments varied among individuals, but 90% of those who participated in the 3 waves had been followed for between 6.5 and 9.5 years. We have not assessed the associations of other variables linked in the literature to frailty, like vitamin D, inflammatory markers, or particular clinical conditions, because we did not want to overfit models.
Conclusions and Implications
Our paper highlights a reduced number of variables that may be of utmost interest to determine the long-term prognosis of patients in the first stages of frailty transitions. Because most of them are also amenable to intervention, it also indicates avenues for intervention, like deprescription, advice on quitting alcohol, and strengthening of upper and lower limbs. It also stresses the importance of registering whether the diagnosis of prefrailty is based on the presence of low grip strength, because this is an indicator of a worse prognosis.
Acknowledgments
The sponsor's role was supporting the analysis and preparation of the manuscript.
References
Kojima G.
Taniguchi Y.
Iliffe S.
et al.
Transitions between frailty states among community-dwelling older people: A systematic review and meta-analysis.
A short physical performance battery assessing lower extremity function: association with self-reported disability and prediction of mortality and nursing home admission.
Effects of changes in number of medications and drug burden index exposure on transitions between frailty states and death: the Concord Health and Ageing in Men Project cohort study.
Risk factors and protective factors associated with incident or increase of frailty among community-dwelling older adults: A systematic review of longitudinal studies.
This work was supported by CIBER, Consorcio Centro de Investigación Biomédica en Red, Instituto de Salud Carlos III, Ministerio de Ciencia e Innovación.