Introduction/Background

As national populations across the world age at an unprecedented rate, disability rates are also increasing (World Health Organization and The World Bank 2011). Prevalence of disability varies greatly by country, however, ranging from 43.4% among persons 60 and older in lower income countries to 29.5% in higher income countries (World Health Organization and The World Bank 2011). High rates of disability can impact individuals’ quality of life and contribute to the health burden of society. This is particularly the case in developing countries, where access to rehabilitation services and treatment programs may be limited (Al Snih et al. 2010).

Extant literature has consistently found a link between Body Mass Index (BMI) and disability among older adults (Davison et al. 2002). Research conducted across racial and ethnic groups suggests that both BMI above the normal range as well as BMI below the normal range are linked to disability (Davison et al. 2002). While the relationship between excess weight and disability is not well understood, the literature suggests that obesity is strongly associated with risk of disability in later life (Al Snih et al. 2007; Chen and Guo 2008; Reynolds et al. 2005) in part because of difficulties with movements, increase in sedentary lifestyle or through disabling pathophysiology (Larrieu et al. 2004). Studies have also shown excess mortality and morbidity among elders with below normal BMI levels. Although some researchers hold that this link is controversial, the association between underweight and ­disability may be due to varying mechanisms, such as lack of physical activity and high risks of falls (Larrieu et al. 2004).

Over the past decade the obesity epidemic has become a pressing public health issue not only in developed countries but also in developing countries such as those in Latin America (Filozof et al. 2001). Transitional countries in Latin America have seen a rise in income, which has been associated with a move away from high complex carbohydrates and fiber towards high-fat diets (Filozof et al. 2001; Uauy et al. 2001). The combination of such dietary changes with increased inactivity contributes to a progressive rise in obesity; Latin American countries are experiencing obesity rates similar to or even in some cases surpassing obesity levels in high income countries such as the US (Monteverde et al. 2010).

Simultaneously, some Latin American countries face significant challenges in undernourishment. Unlike high income countries such as the US, some Latin American countries simultaneously have significant sub-populations that are underweight and undernourished. Rates among underweight elders are particularly high in rural areas of Latin American countries. For example, in rural Mexico over 47% of adults aged 50 and over are undernourished and underweight, compared to 21.3% in urban areas (Wong et al. 2007).

Furthermore, Latin American elders are living in a mixed epidemiologic regime. Specifically, current cohorts of older adults in Latin America are facing rising prevalence of chronic conditions, but infectious diseases continue to impact sub-groups of older adults in many Latin American countries. Recent data indicates that while currently persons in Mexico have lower rates of disability than the US, this may be due to higher infant and childhood mortality levels in Mexico, where only the “fittest survive” (Wong et al. 2010). This suggests that older adults currently living in Mexico are more selected, and possibly more “robust” elders compared to those in the United States (Wong et al. 2010). Additionally, because the two countries are at different stages of the epidemiological and lifestyle transition, elders in the United States are likely to have been exposed for longer in the life cycle to risk factors associated with disability, such as obesity and chronic diseases (Wong et al. 2010). Thus we postulate that it is likely that the effect of BMI on disability will vary between Mexico and the United States. Our hypothesis is that given the stage of the epidemiological transition and the characteristics of the current older adults in Mexico compared to the United States as mentioned above, the overall impact of body weight on disability progression will be lower in Mexico compared to the United States. We hypothesize that both above and below normal BMI will impact disability outcomes in both countries. However, because the nature of disability is likely to be different in the US and Mexico, we expect to see different pathways towards disability in the two countries.

This chapter explores the impact of BMI on the transition to disability among older adults living in two countries at different stages of the epidemiologic and lifestyle transitions. We begin descriptively, examining differences in obesity and disability rates cross-sectionally in Mexico and in the United States. We then model transitions to disability or mortality incidence at 2-year follow up, with a specific focus on the impact of being underweight, normal weight, overweight or obese on the outcomes.

Methodology

Data Sources

In order to compare across countries, this analysis took advantage of the high comparability of the Mexican Health and Aging Study (MHAS) and the Health and Retirement Study (HRS). The MHAS and HRS are large nationally representative panel studies in Mexico and the United States, respectively. The MHAS used a multistage cluster sampling methodology that randomly selected households in Mexico with at least one individual aged 50 or older. There are currently two waves of MHAS data: 2001 and 2003. For a more detailed description of the study see MHAS (2004) and Wong et al. (2006).

The HRS uses a multi-stage national area probability sample of households in the United States with persons over age 50 (see HRS 2008 for more details). The HRS has been conducted biannually since 1992. In order to match to the years of the MHAS sample, the analyses used data from the 2000 and 2002 waves of the HRS, using the dataset prepared by the RAND center (RAND 2010). This is a user-friendly dataset, which assembles each of the HRS waves into one comprehensive dataset (see RAND 2010 for more details).

Whereas the HRS provides data on persons that become institutionalized, the MHAS does not include this information. In Mexico most long-term care is provided by family members, and less than 0.4% of the population resides in nursing homes (INEGI 2000). It is likely that in the MHAS the “loss to follow-up” category captures those people who moved to nursing homes. We addressed this difference in datasets by limiting the baseline sample of the HRS to community-dwelling respondents only. For the follow-up analyses we grouped individuals who moved to institutions and were lost to follow-up into the same category in the U.S., and we included the category lost to follow-up in Mexico. This convention allows us to conduct a complete analysis of transitions in both countries.

Additionally, the HRS sample was further reduced to include only persons that identified as non-Hispanic white and who reported being born in the U.S. Because both the HRS and MHAS interviewed age-eligible respondents and their spouses regardless of age (HRS 2008; MHAS 2004), we only include persons aged 51 and older for the HRS and to make it comparable, we included adults 52 years and older for MHAS, at baseline. The baseline sample size for the HRS was 13,404 persons aged 51 and older in 2000 and 11,837 persons aged 52 and older at the 2001 baseline of the MHAS sample.

Measures

The MHAS and the HRS are highly comparable datasets and unless noted otherwise, measures were similar across the two studies. Disability was assessed using a ­modified version of Katz’s Activities of Daily Living (ADL) scale (Katz et al. 1963). Both ­surveys asked respondents: “Because of a health problem, do you have any difficulty…”. Measures of ADL limitations included: bathing, toileting, transferring into/out of bed, walking, and eating. These five measures have been used by other researchers examining disability in the HRS and MHAS (Hayward et al. 2010). Each of the five measures was dichotomized in disabled/non-disabled. Respondents were coded disabled in the disability measure if they answered “yes” or “can’t do” to having difficulty in performing the activity. If the respondent answered “don’t do the activity” then the response was coded as missing. However, if they answered “don’t do” but also reported receiving help performing the activity, they were coded as disabled.

Independent Variables

All independent variables were taken from the baseline survey of the databases (HRS: 2000; MHAS: 2001). Our focal independent variable for this study was weight. We used Body Mass Index (BMI) to examine respondent’s weight, which was calculated by dividing weight (kilograms) by height (meters) squared. For all regression analyses we grouped BMI into the following categories: <18.5  =  underweight, 18.5–24.9  =  normal (reference group) 25–29.9  =  overweight and BMI ³30  =  obese.

We included several covariates that could be related to the disability outcome, including area of residence (rural versus urban), using rural as the reference category. This variable was measured slightly differently across countries. For the US, the residence variable was based on the 1993 Beale Rural-urban Continuum Codes (for more information, see HRS 2010), which created urban, suburban and ex-urban categories for US residences. We collapsed all urban and suburban areas together and coded ex-urban areas into rural areas. For Mexico, we based the residence variable on the four-category locality size measure available in the MHAS. We considered a community residence with 100,000 people or more the cut-off point for urban; the other three categories were coded rural.

The models also controlled for wealth and having access to health insurance at baseline. Wealth was measured using household’s net worth of homes, businesses, rental properties, capital, vehicles, as well as other debts and other assets. Health insurance was measured by whether the respondent reported having at least one health insurance, regardless of type of insurance. In addition, we created two indices of health conditions for chronic and acute conditions at baseline. Chronic conditions were measured at baseline using questions that asked respondents whether a doctor had ever diagnosed them with diabetes, arthritis or cancer. In order to ­measure the association of recent acute events with disability, we considered whether the respondent had a stroke or a heart attack in the 2 years prior to baseline. For both the MHAS and the HRS stroke was measured by using the question “Has a doctor or medical personnel ever told you that you have had a stroke”. A follow-up question provided information about the date of the most recent stroke; we considered only the acute events that occurred in the 2 years prior to baseline (MHAS 1999–2001 and HRS 1998–2000). This same method was used to determine heart attack or myocardial infarction for the MHAS. The HRS, however, asked respondents ­specifically if in the last 2 years they had ever had a heart attack or myocardial infarction (the question does not specify a doctor diagnosis) and so only this question was used for the HRS. Finally, additional control socioeconomic and demographic variables included age (continuous), gender (reference category: male), marital status (categorical: married and union – reference category – widowed; and single, separated, or divorced), and education (continuous).

Statistical Methods

We begin by presenting sample descriptives stratified by BMI and by country. Next we provide information on the prevalence of disability by country as well as the prevalence of obesity within disability categories.

Then, in order to estimate the impact of obesity across states of physical disabilities between time 1 and 2 in both countries, we take persons reporting no ADL limitations at time 1, and use a series of multivariate models to estimate the probabilities of outcome at time 2. The time-2 outcome is measured by a four-category variable at time two: (1) No disability at follow-up, (2) one disability at follow-up, (3) two or more disabilities at follow-up and (4) death between the first and second wave. Multinomial models were run to determine the likelihood of moving to one of the follow-up categories over 2 years (with reference group: no disability at time 2). We first considered each country separately. We then combined the panel surveys from the two countries and included a country indicator (reference group: United States) to identify significant differences across countries.

The estimators provided by each multinomial model show the relationship between the outcome variable across categories of the independent variables and may be interpreted as a relative risk ratio compared to a reference category (Hilbe 2009). To facilitate the interpretation of results, we also present figures of the estimated probabilities of the outcome variable based on the multivariate models. We estimated the predicted probabilities of each outcome category at follow-up broken down by age. Finally, a post-estimation test comparing across outcome categories was conducted. Post-estimation tests provide a likelihood ratio test for each variable in the model (Long and Freese 2006). This serves as a convenient tool to easily test whether a variable (such as BMI category) significantly affects the likelihood of one outcome category versus the other.

Results

Descriptives: Standardized

Table 6.1 shows the baseline prevalence of obesity and disability by country, as well as the prevalence of disability by each of the four BMI categories (underweight, normal weight, overweight, obese). Due to the difference in age structures across the two countries, all data in Table 6.1 are age standardized using the weighted average of the population distribution of both countries as the standard.

Table 6.1 Age-standardized prevalence of ADL limitations at baseline
Full size table

The data show that obesity (BMI  ³  30) rates are significantly higher in the United States (23.4%) than in Mexico (20.5%). In addition, the disability rates (having at least one ADL limitation) are significantly higher in the United States (11.5%) compared to Mexico (10.4%). While disability rates were higher in the United States for each of the four BMI categories, this difference is particularly notable among obese respondents. Whereas 16.6% of obese respondents in the United States reported at least one ADL limitation, only about 9.6% of obese respondents in Mexico did. All differences were significant at p  <  .01.

Disaggregating further, the information on disability indicates that these cross-country patterns differ by levels of disability. For instance, among persons that were underweight, the prevalence of having one ADL limitation was nearly ­double in the United States compared to Mexico (11.1% vs. 5.7%). However, the prevalence of having two or more disabilities among those underweight was significantly lower in the United States compared to Mexico (9.6% vs. 13.4%). Among respondents that were overweight or obese, the country comparison patterns are only significant for one ADL limitation. Prevalence of having one disability was higher in the United States than in Mexico (5.3% versus 4.0% for those overweight and 8.5 versus 2.6% for those in the obesity category). These differences were significant at p  <  .01.

Descriptives: Bivariate

Table 6.2 provides descriptive information on baseline disability by each independent variable for Mexico and the Untied States. Consistently disability rates are significantly higher in the United States than in Mexico. There are two areas however, where the reverse is the case. This includes the oldest-old age category as well as the health conditions. Among persons aged 80 and older in Mexico, over one-third (35%) reported being disabled, compared to 28.8% of those oldest old in the United States. Additionally, among respondents reporting no chronic conditions (diabetes, cancer, arthritis), 4.0% reported disability in the United States, compared to 6.0% in Mexico. Persons reporting one or two acute condition also had higher rates of disability in Mexico compared to the United States, although these differences were not statistically significant across countries. Across all other categories, the United States reported higher prevalence of having at least one ADL limitation.

Table 6.2 Descriptive statistics of disability at time 1 by characteristics, by country
Full size table

Within each country, there are similar patterns of disability by main socioeconomic variables. Disability is higher among women than among men, and in both countries there is a gradient with age; the higher age categories have higher disability prevalence. Education in both countries also shows a gradient, where those with higher levels of education have lower rates of disability. Similarly, both countries show a decline in disability with higher assets, and higher prevalence of disability among those living in rural areas of the country. Finally, those reporting acute or chronic conditions had higher rates of disability compared to those without acute or chronic conditions in both the United States and in Mexico.

Table 6.3 provides a breakdown of each of the five ADL disabilities by BMI categories. Among obese respondents, the United States sample had significantly higher prevalence of walking, bathing and toileting disabilities. Differences in disability prevalence rates were less pronounced among the overweight respondents, although prevalence rates were significantly higher for bathing and toileting in the United States compared to Mexico. While the pattern appeared reverse for the underweight category (higher prevalence rates in Mexico compared to the United States), none of the differences were statistically significant between the two countries.

Table 6.3 Prevalence of disability by BMI categories, disability component, and country
Full size table

Multinomial Models

Table 6.4 presents results from a series of multinomial models predicting going from zero ADLs at baseline to either one ADL, two or more ADLs, or death at follow-up 2 years later. Persons that remained without any limitation serve as the reference category. Model 1 presents results for the United States sample only, Model 2 presents the results for the Mexico sample only, and Models 3 and 4 show the two combined samples results. For the ease in interpretation, Table 6.4 presents only the results of the focal independent variables (BMI categories and country) as well as the interactions between the two variables. Full multinomial model results are available upon request. The multinomial models provide relative risk ratios (RRR), which may be interpreted as the relative risk of one category compared to the reference category (Long and Freese 2006).

Table 6.4 Multinomial model predicting disability outcome at time 2, HRS and MHAS
Full size table

Table 6.4 Model 1 (United States sample) shows that obesity (as compared to otherwise similar respondents of normal weight), is significantly predictive of disability incidence in the United States. Obese respondents have a higher risk of moving from zero disabilities to one disability or to two or more disabilities over the 2 year time-span. Conversely, the model indicates that obesity has a protective effect on the probability of death in the United States. Being underweight was only predictive of higher risk of death at 2 year follow-up.

These relationships differ for Mexican elders (Table 6.4 Model 2). The relationship between obesity and disability is weaker for Mexico compared to the United States. The relative risk ratios of obesity on ADL outcomes are lower, and only the negative effect of obesity on incidence of one ADL limitation is statistically significant. The relationships of obesity to the other outcomes were in the same direction as in the United States model, but were not significant. Being underweight at baseline was not statistically significantly associated with any of the disability outcomes in the Mexican sample.

Table 6.4 Model 3 combines the data from both countries to create a pooled sample model with a country identifier variable. The results from this model show that these differences in associations between the two countries are statistically significant only for the outcome categories of death and loss to follow-up. Finally, Model 4 adds an interaction variable to the previous models, interacting BMI categories with country. The interaction effects show that the relationship between being underweight and death at 2 year follow-up is significantly bigger (steeper) for the United States than for Mexico.

Predicted Probabilities

Figures 6.1, 6.2, and 6.3 graphically show the predicted probabilities of the disability outcome categories across countries and break these down by gender and educational levels. Figure 6.1 presents predicted probabilities based on the multinomial models for one ADL, two or more ADLs and death in Mexico (Fig. 6.1a) and the United States (Fig. 6.1b). The probability of death is higher in the United States than in Mexico for all four BMI categories, especially for the underweight category. The predicted probability curve for death appears relatively flat in Mexico, compared to the United States. While for both countries the highest probability of death was in the underweight category, the probabilities do not vary as much in Mexico for the other BMI categories. Whereas the shape for BMI and mortality in the United States resembles an L shape, the shape is relatively flat for Mexico.

Fig. 6.1
figure 1_6

(a, b) Predicted probabilities of one ADL, two ADLs or more, and death at time 2, by country. Notes: See Table 6.4

Full size image
Fig. 6.2
figure 2_6

(a) Predicted probabilities of death at time 2, by country and gender. (b) Predicted probabilities of one ADL and two ADLs or more at time 2, by country and gender. Notes: See Table 6.4

Full size image
Fig. 6.3
figure 3_6

(a, b) Predicted probabilities of one ADL and two ADLs or more at time 2, by country and education level. (c) Predicted probabilities of death at time 2, by education level and country. Notes: See Table 6.4

Full size image

The predicted probabilities of ADL limitations appear to be in a U-shape pattern for the United States, with the highest probabilities for transitions to either one ADL or two or more ADLs appearing in the extreme BMI categories (underweight and obese). This U-shape is more pronounced for the probability of one ADL incidence. In Mexico these patterns are quite different. In the Mexican sample the pattern for predicting one ADL appears slightly linear and for two or more ADLs it appears relatively flat. Mexican elders that are obese have the highest predicted probability of having an ADL limitation or two or more ADL limitations, which is similar to the United States results. However, underweight persons have a lower predicted probability of one ADL compared to normal weight and overweight respondents. Overall, the probability curve for two or more ADLs in Mexico is relatively flat compared to the United States. This graphically depicts the findings reported in the multinomial models; the association between obesity and ADL limitations is less strong for Mexico than for the United States.

When examining these effects by gender (Figs. 6.2a, b), the predicted probability figures show that the effect of gender is similar in the United States and in Mexico in that both countries show a clear gender gap for every outcome. In both countries the probability of death is higher among men than women, regardless of BMI status (Fig. 6.2a). This gender gap is particularly pronounced in the underweight category in the United States. On the other hand, the incidence probability of one ADL or two or more ADLS is higher for women in both countries (Fig. 6.2b). In both countries the gender gap is consistent for each of the four BMI categories and for both disability outcomes.

Figure 6.3a–c break the predicted probabilities down by low and high education level achieved. These two categories were used to compare across two countries with vastly different distribution of educational achievement. While high education appears protective for outcomes in both countries, this differs somewhat by outcome. In Mexico (Fig. 6.3a) the education gap is greater for the more severe ADL outcome (two or more ADL limitations), whereas the gap does not differ much by ADL severity in the United States (Fig. 6.3b). For the mortality outcome (Fig. 6.3c), in Mexico the education gap is relatively small across all four BMI categories. In the United States, however, education results in a consistent and relatively wide gap in mortality, where low education predicts higher probability of mortality at 2 year follow-up.

Post-Estimation

Post-estimation results assess statistically the discriminatory power of the BMI categories to differentiate between two given categories of the outcome variable. In the Mexican sample, there were no significant contrasts for the underweight category. The obesity category only had an effect on two contrasts: one ADL and loss to follow-up/nursing home, and one ADL and no ADLs in the Mexican sample. However, in the United States post estimation showed that obesity had a significant effect on predicting one ADL compared to death, no ADL compared to one ADL, two ADLs compared to death, no ADL compared to two ADLS and no ADLs compared to death. The only output differences that were not significant for the United States were one ADL and two plus ADLs. The underweight variable had a significant effect on the contrasts of no ADL limitation and death as well as loss to follow-up/nursing home and death only. Overall, these results indicate that the body mass indicator has clearer and stronger differentiating effect across disability categories in the United States compared to Mexico.

Discussion

This chapter explored the association between BMI and disability across older adults in a developing country (Mexico) and a developed society (white non-Hispanic population in the United States) as a benchmark. Results show that both obesity and disability were more prevalent in the United States than in Mexico. This echoes previous research indicating that disability rates are higher in the United States than in Mexico (Wong et al. 2010). Our findings that the association between obesity and disability is stronger in the United States than in Mexico supports the theory that the current cohort of older adults in Mexico are highly selected in terms of survival, thus the impact of lifestyle changes on health outcomes (such as disability) are weaker for Mexico compared to a less-selected groups of older adults (non-Hispanic whites in the United States).

Based on this selective survival hypothesis, we anticipated that the prevalence of ADL limitations would be higher among overweight and obese persons in the United States. The breakdown in prevalence of the five types of ADL disabilities confirmed this, where overweight and obese elders in the US had higher prevalence of several ADL limitations compared to Mexico. In addition, we expected the gap between the two countries to be larger at older ages. However, our findings show that the difference actually reverses for those aged 80 and older. It is possible that in this oldest-old category we are comparing the fittest in both countries, and possibly the fittest in the United States are healthier and less disabled than those in Mexico. Further research is needed to explore this issue in more detail.

While our finding of lower rates of disability in Mexico echoes that of previous research using the same samples, it is important to note that other researchers have found that Mexicans aged 65 have higher prevalence of ADL disability than older Mexican Americans living in the United States (Patel et al. 2006). This difference in findings is likely due to the difference in age groups and a very different United States comparison group, which makes comparisons to the current study results difficult.

The multinomial models and predicted probability figures show that there are differences between Mexico and the United States in ADL incidence at 2 year ­follow-up by BMI category. The patterns for ADL limitations appear as a U-shape in the United States, where risks for ADL limitations are highest among underweight and obese American elders. For Mexico, however, the BMI-disability gradients are relatively flat for the probability of two or more ADL limitations, and slightly linear with a low slope for the onset of one ADL. Previous research echoes these findings. In recent research examining hazard ratios of ADL limitation in the United States, researchers found a U-shaped curve, with lowest hazard ratios for disability at BMI of 24, and steep increases in risk of disability at higher and lower BMI measures (Al Snih et al. 2007). On the other hand, a study examining elders in Mexico City, Mexico, found a relatively flat association between BMI categories and any ADL limitations (Al Snih et al. 2010). Using cross-sectional data, the researchers found that ADL limitations increased slightly with higher BMI levels, but overall the curve remained relatively flat. These patterns reported in the United States and in Mexico support the findings in our analyses using national samples and two-wave panel data for the two countries.

Additionally, our models indicate that obesity at baseline was associated with lower risk of mortality at follow-up for the United States. While the effect was in the same direction in Mexico, the effect was weaker for Mexico than for the United States (though this difference in effect was not statistically significant) and it was only significant in the United States. The higher relative risk of mortality with excess adiposity found in the United States analyses echo previous research finding a protective effect of obesity on mortality among older adults (Grabowski and Ellis 2001). This differs from the increase in mortality risk among obese younger populations. Explanations for this finding vary but include suggestions that the BMI cut-off point is too low for older adults, that there are flaws in study designs or that BMI is not an accurate measure of adiposity in the older population (Al Snih et al. 2007).

The models also indicate a large and statistically significant difference between Mexico and the United States in the probability of death among underweight persons. This finding may be due to a difference in the health profile of older underweight adults in Mexico compared to the United States. It is possible that underweight older adults in Mexico have lived a large portion of their lives being underweight, whereas being underweight may be more likely to be associated with illness and unexpected weight loss in the United States non-Hispanic white population. This hypothesis is supported by the finding that in our sample of underweight elders in the United States, 75% have at least one chronic condition; in contrast, only a little over 1/3 (36%) of Mexican underweight elders had at least one chronic condition. This suggests that the profile of underweight elders is different in Mexico than in the United States. However, this is only one possible explanation for this relationship, and further research should be done on this specific group of vulnerable elders.

Both countries show a clear disability gradient by gender and by education. However, the education gap is wider in Mexico for the more severe ADL limitation outcome, and the gap is wider also in the United States for the mortality outcome. Because these two countries have vastly different education systems and the cut-off points for high/low differ, it is unclear what the underlying relationships might be. It does, however, suggest that education plays a role in the relationship between BMI categories and incidence of ADL limitation or mortality.

Limitations

These analyses had several limitations. First, the models relied exclusively on self-reported measures for the dependent variable (ADL disability) as well the measures used to create BMI (height and weight). While research has established a strong concordance between self-reported ADL limitations and observed limitations (Reuben et al. 1992), it is possible that there may be a bias due to self-report and that this bias may differ across countries. Self-reported height and weight has been shown to produce somewhat lower estimates of BMI than when directly measured among older adults (Kuczmarski et al. 2001), which would skew the BMI measure in our study downward. Our results apply only to community-dwelling populations of older adults at baseline. To make our samples comparable across the two countries, our baseline sample for the United States excluded population that live in institutions and this may bias our cross-sectional results. However, for the transition analyses we take this limitation into account by including persons that moved into nursing homes or were lost to follow-up. Additionally, our analyses are based on comparative research, which has serious limitations (Angel et al. 2008). While the two survey instruments are highly comparable, it is important to note that the two independent surveys were collected at different times in two different countries.

Conclusions

The results from these analyses suggest that obesity affects the physical disability progression differently according to the stage of the epidemiological transition that societies are undergoing. Mexico is currently at earlier stages of the epidemiologic and demographic transitions compared to the United States, but it is likely that as Mexico advances in the transition, the population will be less selected in terms of survival, and if patterns of lifestyle changes continue the path followed by the United States, the Mexico population will experience an increase in obesity levels. Furthermore, the results from these analyses show that being underweight is significantly associated with higher risk of death in the United States, and additionally show that increases in obesity result in increases in disability at 2 year follow-up. This underscores the urgency for public health and policy intervention in obesity. Preventing the obesity epidemic from spreading in Mexico and spreading further in the United States could reduce disability levels, resulting in positive impact for individuals as well as for society overall. Our work suggests the need to continue monitoring changes in obesity and disability levels in these two countries in ­particular, and in general across societies as they continue to age, and to move along the epidemiological transition.