A frailty index to predict the mortality risk in a population of senior mexican adults

Frailty in the elderly can be regarded as nonspecific vulnerability to adverse health outcomes, caused by multiple factors [1]. Although in the literature there are many other operationalizations of a frailty definition [2‐7], two of them have been widely used. One approach, proposed by Fried et al, considers frailty a distinct clinical syndrome not synonymous with either co morbidity or disability [8]. Under this perspective, frailty is present when an individual has three of the following criteria: unintentional weight loss (10 lbs in past year), self-reported exhaustion, weakness (grip strength), slow walking speed, and low physical activity. Additionally, an intermediate status (prefrailty) is also identified when the individual presents one or two of the above criteria. The other approach, proposed by Mitnitski and Rockwood [9], considers frailty as a continuum of accumulation of deficits (symptoms, signs, diseases, and disabilities) that are combined in a frailty index score, reflecting the proportion of potential deficits present in a person [10].

Apart from the conceptual differences, the two approaches differ in terms of measurement conditions. Fried's approach requires direct measurement of the proposed criteria by health personnel, whereas Mitnitski and Rockwood's approach relies on self report of the presence of deficits. Yet, both approaches have been successfully used in identifying elderly vulnerable to a range of adverse health, outcomes including mortality [1, 8].

Three studies have compared both approaches in the same study sample in identifying elderly at high risk of death [11‐13] and have found correlation between them, with the deficit accumulation approach predicting mortality better [11, 12].

In this paper, we present an analysis of a two year follow up study of Mexican elderly using the deficit accumulation approach to predict mortality.

Figure 1 summarizes the characteristics of the analyzed sample. A total of 4872 persons aged 65 and older were interviewed in the MHAS 2001 wave. Of them, we excluded 449 persons (9.2%) because they were not directly interviewed. On average they were 2.6 years older and had a three-fold higher mortality compared to those directly interviewed. In addition, 196 (4.4%) of the 4423 interviewed participants were lost to follow-up. Compared to those not lost to follow-up, not followed-up participants were slightly more educated (1 school year more on average), and had a higher income (2.3 times on average). Finally, after excluding 145 of those followed-up (3.4%) because of missing frailty index data, we analyzed information on 4082 participants (92.2% of those directly interviewed). Analyzed participants had an average age of 73 years (range 65 to 105), with 52.5% being women, both of these figures being very close to the Mexican population aged 65 years and older. On average, analyzed participants were followed-up for 710 days (standard deviation = 111 days) and 279 of them died during the follow-up.

Coding of the health deficits composing the frailty index are shown on table 1, along with their sample weighted mean value, which can also be interpreted as the prevalence of a particular deficit. The most prevalent were vision and hearing problems, leg pain on walking, high blood pressure, falls in the past two years and depressive symptoms with average values between .4 and .5. On the other hand, Cancer and most health problems before age 10 showed the lowest prevalence in the elderly Mexican population with average values below .02.

The constructed frailty index had an average value of .16 in the target population (standard deviation = .11), ranging in the sample from 0 to .65 and showing a right-skewed distribution (Figure 2).

In addition, the mean frailty index increased with age between the ages of 65 to 89 and then decreased in those aged 90 and older (Figure 3). This age effect was more clearly seen in men. In fact, women showed significantly higher mean frailty index values than men in the age groups younger than 80 years.

The following five frailty index levels were obtained by the minimum variance method (with percentages of the Mexican elderly population shown in parenthesis): .00-.07 (17.4%), .07-.14 (30.8%), .14-.21 (24%), .21-.35 (21.4%), .35-.65 (6.5%). Figure 4 shows the cumulative percentage of persons dying during the follow-up by frailty level for an average Mexican with respect to age (73 years) and gender (52.5% females). Over the follow-up period, mortality in older Mexican adults increased with the frailty index level, especially in those with frailty index levels between .21 to less than.35 and those with frailty index levels between .35 and .65, reaching at 2 years of follow-up approximately 17% and 21%, respectively.

These findings were confirmed in the Cox proportional hazards models (Table 2). In all participants those in the two highest frailty levels have age and gender adjusted death hazards 2.2 and 6.45 times higher than the observed in the lowest frailty level. Women experience a 33% lower instantaneous risk of death than that observed in men and each year of increase in age is associated with a 5% increase in the risk of death. Cox models fitted separately in each gender show that in men there is a statistically significant increase in the age adjusted death hazard in both those with frailty index levels between .21 to less than.35 and those with frailty index levels between .35 and .65. In contrast, women show only a statistically significant increase in the age adjusted death hazard in the highest frailty index level (.35-.65).

Finally, figure 5 shows the prevalence of frailty index levels associated to increased mortality hazards in the Cox models by age group in the non-institutionalized Mexican elderly population. Both frailty levels show an increasing prevalence with age, especially after age 70. Elderly Mexicans with frailty index between .21 and less than .35 represent 19.7% of those aged 65-69 years and 29.8% of those aged 85 and older. More marked increases in the prevalence of persons with frailty index values .35 or higher are seen. These frailty index values are seen in 4.3% of those aged 65-69 years and 17.7% of those aged 85 and older. Thus, the proportion of Mexican older adults with frailty index levels associated to increased mortality (.21 and higher) represent 24.0% of those aged 65-69 years and 47.6% of those aged 85 and older.

This paper presents the results of an analysis of a frailty index, based on self report accounts of deficit accumulation and developed by Mitnitski and Rockwood [9], as an independent predictor of mortality in a probabilistic sample of non-institutionalized Mexican older adults, aged 65 and older. 92% of analyzed participants were followed up, for an average time of almost 2 years. This implies that we evaluated mortality over a relatively short period after the initial interview.

Interviews were conducted face to face by trained personnel and only those directly administered were analyzed here. We did not analyze data from 449 of the 4872 initially interviewed participants (9.2%) because a proxy had answered a questionnaire with insufficient data to construct the frailty index. These participants had a higher mortality than those directly interviewed and were likely to be very ill. Therefore, excluding them from our analysis could have weakened the association we found between the frailty index and mortality. In addition, the prevalence estimates of high frailty levels we report may be an underestimate of the values we could have found had we not excluded those with proxy questionnaires.

In spite of the above caveats, our study replicates previous findings from secondary analysis of similar datasets from North American and European developed countries [17], as well as from China [21, 22], in institutional and community elderly using roughly equivalent versions of the deficit accumulation frailty index. In our analysis the frailty index showed a skewed distribution which, as previously found by Rockwood et al [23], could be fitted by a gamma distribution and a maximum value of .65 was observed, as others have also found [24, 25].

As has been previously reported in other settings[9, 15, 17, 26], both the mean frailty index and the prevalence of high frailty levels were found to increase with age in our sample of Mexican elderly, with some leveling off at very advanced ages possibly explained by selection of robust individuals at those ages [27, 28]. In addition, we found that mean frailty levels were higher in women than in men before age 85, as other studies have also found [29]. The fact that women seem to accumulate more deficits, as measured by the frailty index, at a faster speed in late life may be explained by their higher rate of objective disability [1, 30], possibly caused by genetic [27] and social factors [31], but also by better reporting of medically diagnosed conditions in women due to their greater access to health care [32].

In our analysis, the categorized frailty index strongly predicted 2-year death, independently of chronological age, in those with a frailty index with a value of at least .21, particularly in men, a finding consistent with previous studies [9, 15, 21, 22, 33, 34].

This concept appears applicable to populations with diverse ethnicity and life styles. However, it would be important to study populations from developing countries with socioeconomic deprivation in order to validate the construct.

Regardless of how frailty is conceptualized and measured, vulnerability towards death has been explained by two biological mechanisms: allostatic load and sarcopenia. Seeman [35] defines allostatic load as the cumulative physiologic toll exacted on the body over time by efforts to adapt to life experiences. As an individual age he or she experiences complex losses in resting dynamics and maladaptive responses to perturbations [36]. This process starts likely early in life and this is the reason why we included serious health problems experienced before age 10 in our frailty index. Participants who reported a serious event during childhood can be defined as survivors. Because of little is known about the effect of early life events on aging course, these findings will mark further research.

Sarcopenia refers to the substitution of muscle by fat tissue, accelerated by physical inactivity, chronic diseases and disability, among other factors and has been found related to several adverse health outcomes [37]. In fact, studies using Fried's frailty definition have found that frail older adults have lower muscular density and muscle mass and higher fat mass than non frail elderly [38]. Beyond to this two biological explanations, it is possible to say that a combination of several factors such a lifestyles, social networks, economic status, emotional life are frequently behind of an specific accumulation of deficits [39]

Women showed a weaker association between the frailty index and the risk of dying over the next two years, as has also been found elsewhere [15, 22, 27]. Whether this is explained by a true biological difference [40] or, as stated above, by differential measurement error of a frailty index based on self report is a matter that needs to be further addressed.

To our knowledge this is the first study providing data on the magnitude of frailty in a Latin-American country and of its impact on mortality. We found that one in four Mexicans aged 65 to 69 years and almost one in two Mexicans aged 85 and older had frailty levels that increased their risk of dying over the next two years. This has obvious health care implications in the light of studies showing that frailty is a dynamic process, characterized by frequent transitions to states of lower frailty and therefore suggesting opportunities not only for preventing but for alleviating frailty [27, 41].

A frailty index based on a simple count of self reported deficits offering an adequate prediction of mortality, both in clinical and epidemiological settings, is a useful tool to deal with frailty and its consequences, especially in developing countries.