Predictors of Frailty Transitions in Mexican Older Adults

Abstract

Background

Frailty is a dynamic state in older adults. Current evidence, mostly in high-income countries, found that improving frailty is more likely in mild states (prefrailty). We aimed to determine the probability of frailty transitions and their predictors.

Methods

Participants were adults aged 50 years or over from the Study on Global Ageing and Adult Health in Mexico during 4 waves (2009, 2014, 2017, and 2021). We defined frailty with the frailty phenotype and we used multinomial logistic models to estimate the probabilities of frailty transitions and determine their predictors.

Results

For the 3 analyzed periods (2009–2014, 2014–2017, and 2017–2021), transition probabilities from frail to robust were higher for the younger age group (50–59 years) at 0.20, 0.26, and 0.20, and lower for the older age group (≥80 years), 0.03, 0.08 and 0.04. Transitioning from prefrail to robust had probabilities of 0.38, 0.37, and 0.35, for the younger age group, and 0.09, 0.18, and 0.10, for the older age group. The probabilities of transitioning to frail and to death were lower for the younger age group and for the robust at baseline; but higher for the older age group and for the frail at baseline. We identified age, disability, and diabetes as the most significant predictors of frailty transitions.

Conclusions

These findings show that frailty has a dynamic nature and that a significant proportion of prefrail and frail individuals can recover to a robust or prefrail state. They also emphasize that prefrailty should be the focus of interventions.

Latin America, the Caribbean, and some Asian regions are seeing an accelerated aging of their population, with the number of adults 65 or older expected to double in the next 30 years (1). Older adults have heterogeneous health states, although some live with full functionality, others live with poor health (2), for instance, frailty. Physical frailty results from a critically dysregulated complex dynamical system (3). Frailty is a state of vulnerability to stressors, with decreased physiological reserve and increased risk to adverse outcomes, including falls, hospitalization, disability, and mortality (4).

Frailty is a dynamic process with transitions between robustness and frailty states, it presents critical transitions where resilience plays a compensatory role; otherwise, it results in progression to a worse state (3). According to a systematic review, 29.1% of older adults worsened and just 13.7% improved their frailty status, whereas 56.5% remained in the same status (5).

Some predictors for frailty progression are older age (6–8), being female (7), increase in body mass index (7), fewer years of education (7), multimorbidity (9), and dependence in activities of daily living (7,8). Conversely, predictors for frailty improvement include better cognitive function (6,7), normal waist circumference (9), higher education (7), and no history of chronic diseases (CDs) (10).

Research into factors predicting frailty transitions is mainly based in wealthier countries. In contrast, evidence is scarce in low- and middle-income countries (LMICs) even though LMICs face a high burden of CDs interacting with persistent social and economic disadvantages (11). Additionally, there are some less-studied predictors, such as alcohol consumption, smoking, physical activity, waist circumference, and falls. Thus, our main aim was to estimate the probabilities of frailty transitions among older Mexican adults through longitudinal data from the Study on Global Ageing and Adult Health. A secondary purpose was to determine the variables that could predict frailty transitions.

Method

Sample

Data come from the 4 waves of the World Health Organization (WHO) Study on global AGEing and adult health (SAGE) in Mexico. A longitudinal, multicenter study, SAGE, is conducted by the WHO across 6 LMICs: India, Ghana, Russia, Mexico, South Africa, and China. The methodological details have been previously reported elsewhere (12). SAGE provided nationally representative groups of adults aged 50 and older. SAGE applied standardized instruments with face-to-face interviews, including household and individual information, as well as anthropometric and functionality measurements.

For the present study, we include information on adults 50 years or over from the 4 waves of the SAGE-Mexico. Wave 1 (baseline) was collected between July and September 2009; Wave 2 between July and October 2014; Wave 3 between August and November 2017, and Wave 4 between May and July 2021. The analytical sample included all individuals 50 years and over with full information in 2 consecutive waves: 2009–2014 (n = 1 648), 2014–2017 (n = 2 020), and 2017–2021 (n = 1 602). (Supplementary Figure 1)

Study Variables

Frailty

We used the frailty phenotype proposed by Fried et al.(4), which covers 5 components: unintended weight loss, exhaustion, low physical activity, slow walking speed, and weakness. Subjects considered frail presented 3 or more of these components; prefrail, 1 or 2; and robust, none (4). We thus expressed categorically the results as 0 = frail, 1 = robust, and 2 = prefrail. Details on each component are reported in Supplementary Tables 1 and 2.

Death was considered a part of the frailty transitions in the follow-up and was recorded using a verbal autopsy.

Predictors

Health-related

Disability was measured using the World Health Organization Disability Assessment Schedule (WHODAS 2.0) (13). It covers 6 domains (12 items): (1) cognition and communication, (2) self-care, (3) mobility, (4) interpersonal relations, (5) life activities, and (6) participation. We obtained a global score on a continuous scale from 0 (no disability) to 100 (full disability). Chronic diseases: We used a list of 9 CDs: diabetes, stroke, cataracts, angina, arthritis, chronic obstructive pulmonary disease (COPD), asthma, depression, and hypertension. The CDs were assessed dichotomously (0 = absence, 1 = presence) (11). Self-rated health (SRH): We created a categorical variable with 3 levels: 0 = good health (good and very good), 1 = moderate and 2 = poor health (bad and very bad). Cognitive function: SAGE measures 5 cognitive performance tests: immediate and delayed verbal recall, forward and backward digit span, and verbal fluency. We created a composite z-score to compare individuals’ measures of cognitive tests. Visual impairment was defined as having severe and/or extreme difficulty with far and near vision (1 = with visual impairment). Central obesity: We defined central obesity according to the International Diabetes Federation cut-points for waist circumference: men >90 cm and women >80 cm (1 = healthy waist circumference). Falls: We defined this variable as having a fall-related injury in the last 12 months (1 = yes). Hospitalization: We used the self-report of being hospitalized at least 1 night in the past year (1 = yes). Health coverage (1 = with health coverage).

Sociodemographic

We used the following variables: categorized age (50–59, 60–69, 70–79, 80+), sex (1 = female), education (years of schooling), marital status (1 = with partner), place of residence (1 = rural), employment status (formal employment or retired = 1), and household income using a continuous index based on household assets, where higher positive values indicated higher incomes.

Social capital was measured at individual level and examined in structural and cognitive dimensions (14). We generated a variable with 4 categories: no access to either structural or cognitive social capital = 0, with access to both dimensions of social capital = 1, structural only = 2, and cognitive only = 3. Living arrangements: We identified individuals “living independently” = 0 (living alone or with a partner); “living with children” = 1 (couple or single parent living with children); “Extended family households” = 2 (households with 1 or more member that are not part of the nuclear unit, ie, outside the 2 previous categories). Loneliness: It was assessed by the question: Did you feel lonely for much of the day yesterday? (1 = yes).

Health-related lifestyle behaviors

Alcohol drinking: We constructed a categorical variable: lifetime abstainer = 0, ever but not past-week drinker = 1, low-risk drinker = 2, and high-risk drinker = 3. Tobacco use: Based on questions about frequency and use of tobacco, we defined 3 categories: never smoker = 0, past but not currently = 1, and current smoker = 2. Nutrition: Fewer than 5 servings of fruit/vegetable are insufficient = 1.

Statistical Analysis

We presented the characteristics of the sample at each wave as means (standard deviation) or percentages as appropriate (Supplementary Table 3). To model frailty transitions, we use the approach of multistate models, in particular, the so-called illness–death model with 4 states, 3 transitory (robust, prefrail, frail), and 1 terminal or absorptive (death) (15). We fitted age-stratified (50–59, 60–69, 70–79, 80+) multinomial logistic regression models to estimate the frailty transition probabilities and analyze their predictors considering the follow-up time between waves (days elapsed). We used the average marginal effects to report the effect of each predictor. We performed statistical analyses using the software Stata 16.1 (StataCorp LP, College Station, TX).

Ethics

Ethics and Research Committees of the National Institute of Public Health in Mexico approved the SAGE-Mexico. All participants were provided with a detailed explanation of the study procedures and signed an informed consent letter.

Results

Transition Probabilities

In this study, older adults of all age groups presented higher probabilities of improving (from frail or prefrail to robust) than worsening (from robust or prefrail to frail). The probabilities of transition from frail to robust for the younger age group (50–59 years) were 0.20 (Period 1), 0.26 (Period 2), and 0.20 (Period 3), for the older age group (80+ years) were 0.03, 0.08, and 0.04. Although the probabilities of transition from robust to frail were 0.02 for older adults of 50–59 years and 0.09 for older adults of 80+ years. Being prefrail at follow-up was the most probable transition regardless of the baseline state, the period analyzed, and for older adults younger than 80 years, with probabilities ranging from 0.35 to 0.61. Older adults with frailty at baseline and 80 years or over had higher probabilities of dying (Supplementary Figure 2).

Predictors of Frailty Transitions in Older Adults ROBUST at Baseline

For all the periods analyzed, the increase in age and disability decreased the probability of remaining robust. Further, reporting poor SRH decreased the probability of maintaining robustness for the first and second periods. A higher score of disability increased the likelihood of transitioning to prefrail for the first and third periods. In all periods, an increase in age and diabetes increased the probability of dying, whereas disability only increased this probability in the last 2 periods (Figure 1 and Supplementary Table 4).

Predictors of Frailty Transitions in Older Adults PREFRAIL at Baseline

For those who were prefrail at baseline, age, disability, and diabetes decreased the likelihood of improving to robust over the 3 periods. Having angina and being female reduced the chance of improving to robust in 2 of the 3 periods. Age and being female were predictors for remaining prefrail. Age and disability were the most consistent predictors for worsening to frail. For all the periods, age and diabetes increased the likelihood of dying. Being female decreased the probability of dying for the first and second periods, while disability increased this probability (Figure 2 and Supplementary Table 5)

Predictors of Frailty Transitions in Older Adults FRAIL at Baseline

For the first and second periods, having hypertension decreased the likelihood of improving from frail to robust. Older age increased the probability of remaining frail. For all the periods, disability increased the probability of dying. However, a higher cognition score decreased the probability of dying for the first and third periods (Figure 3 and Supplementary Table 6).

Discussion

In this study, we analyzed the probability of transitions between frailty states and their predictors. Overall, we found that the probability of improving was higher than the probability of worsening; these probabilities were higher for the younger age group. Prefrailty was the most probable transition at follow-up for all baseline states. In relation to the predictors, some showed a persistent effect across the waves studied, mostly for prefrail older adults. Predictors were less consistent in the frail baseline subgroup. Our results indicated that age, disability, and diabetes were the strongest predictors of frailty transitions.

According to a systematic review that included studies with a wide range of follow-up intervals (from 1.5 to 10 years) (5), only 2.0% of frail participants became robust, which is similar to what we observed in the older age group (80+ years), but smaller than what we observed for younger ages, especially the age group of 50–59 years (20%, 26%, and 20%, for each period, respectively). Our probabilities of improving from frail to robust at younger ages (<80 years) are similar to those reported by Rodríguez-Laso et al. (8), who found a 15% improvement in a Spaniard cohort of 65+ older adults. Also, there is evidence that the length of follow-up intervals affects the probability of recovery. A scoping review that divided the results of frailty transitions by the follow-up intervals showed that short and intermediate intervals (≤6 years), similar to our study, had higher rates of improvement compared with long intervals (>6 years) (16). Thus, the length of our follow-up intervals might partially explain our probability of recovery from frail to robust.

Evidence on predictors of frailty transitions includes older age, female, CDs, and disability (6,16,17), which is consistent with our findings. Prior research has suggested that determinants of frailty transition (for instance, CDs and disability) may be associated with physiological reserve depletion and chronic inflammation (3,16)

Our results showed that increased disability scores corresponded with reduced chances of remaining robust at follow-up, as well as the highest probability of frailty and death. Disability, as measured by physical function, has been associated with transitions to frailty, particularly worsening to prefrail and mortality in robust people (7).

In understanding the association between diabetes and frailty, several mechanisms have been studied, one of them being insulin resistance which appears to have a great influence on skeletal muscle performance (18). Diabetes has been linked with a decrease in the probability of improving to robust in prefrail individuals and an increasing risk of death in robust older people (6,7), which is in line with our results (16).

Aging and cardiovascular diseases have been associated with worsening and a lower chance of improving frailty state. They are associated with a pro-inflammatory state, consequently, a decrease in muscle mass and strength, low physical activity, and depression, all of which are related to frailty (7,9,19). In our study, angina, hypertension, and cardiovascular diseases were associated with the frailty transitions, decreasing the probability of improving to robust for the prefrail and frail subgroups.

Sex was a consistent predictor for the prefrail subgroup, but without differences in frail older adults. Other studies have reported that being female is associated with a worsening robust and a lower risk of death (7). Evidence suggests that women experience higher rates of prefrailty and frailty but lower mortality rates (20). Differences in biological, psychosocial, and behavioral factors might be related to the sex differences in frailty, for instance, differences in immunosenescence, hormones, CDs, skeletal muscle changes with age, social vulnerability and coping strategies, stress perception, and lifestyle behaviors (20).

According to our data, it is possible that better cognitive performance could reduce the risk of mortality in frail individuals. Other studies have reported that better cognitive function decreased the probability of worsening and result in an underestimation cognitive impairment increased the probability of worsening (6,7).

Our results showed that COPD and waist circumference were associated with frailty transitions; however, these results were less consistent and observed only for 1 period and/or 1 transition. More studies are required to understand the effect of those variables on frailty transitions.

Our study has some limitations. First, the analytical sample differed from the excluded sample in some characteristics (Supplementary Table 7). These variables are associated with the worst health status and we could be underestimating the transition probabilities of frailty. Nevertheless, we did not observe differences in the frailty categories and other important covariates. Second, the interval among waves does not consider transitions within these intervals, and there is some evidence that changes in frailty status are highly frequent (17). We might underestimate some transitions at each follow-up, especially among individuals with faster health deterioration who became frail and died during the follow-up interval. Consequently, we might lose the opportunity to observe “frailty” before “death.” This discretely lengthened data collection results in an underestimation of subjects who transit to a worse frailty state, and partially explains why participants with frailty had a high probability of improving. One strength of our study is that we were able to compare 3 follow-up periods and verify our probabilities of transitioning among frailty states. Also, we explored a wide number of variables to identify the most important predictors of frailty transitions in our models.

In conclusion, our findings show that frailty has a dynamic quality and that a substantial amount of prefrail and frail individuals can recover to a robust or prefrail state. They also emphasize that prefrailty should be the focus of health and socioeconomic interventions for older adults.

Funding

SAGE is supported by World Health Organization and the US National Institute on Aging through Interagency Agreements (OGHA04034785, YA1323–08CN0020, and Y1AG100501) and a competitive grant: R01AG034479.

Conflict of interest

None.

Author Contributions

Study concept and design: A.R.A., B.M.E., E.G.P., A.S.R.; formal analysis: A.R.A., E.G.P., A.S.R.; writing, review and editing: all the authors. All the coauthors approved the final version.

References