Recovery from Physical Limitations among Older Mexican Adults

Abstract

Objectives:

This paper examines the key determinants of the likelihood of recovery from a physical disability among older adults.

Methods:

Data come from the Mexican Health and Aging Study (MHAS), a national sample of adults born in 1951 or earlier, including a baseline survey in 2001 and follow-ups in 2003, 2012 and 2015. At baseline, we divided our sample of older adults aged 60+ by dimensions of physical limitations (ADLs, IADLs, mobility) and classified respondents as having physical limitations in zero, one, two or three dimensions. Each respondent was then categorized as “same”, “worse”, “improved” or “died” depending on the number of physical dimensions with a limitation in a 2-year span (2001–2003) and again, separately, in a 3-year span (2012–2015). We then used a multinomial logistic regression to analyze the relative risk of transitioning from one category to another.

Findings:

Around 21% of our sample exhibited some recovery in 2003 and around 20% recovered in 2015. Age, gender, poor self-rated health, depression and some chronic conditions were significant for shifting the relative risk from staying the same to getting worse, dying or even improving.

Conclusions:

Disability from a physical limitation is a reversible and dynamic process. Our results reflect the importance of considering the dimensions of physical ability while analyzing recovery, and illustrate that the presence of a chronic condition or depressive symptoms does not necessarily imply permanent disability.

1. Introduction

The dynamics of aging in developing countries has begun to receive more research attention as reliable longitudinal survey data continues to become increasingly available (Olaiz et al., 2003). Previous research has found that the development of physical disability seems to follow a progression (Harris, Kovar, Suzman, Kleinman, & Feldman, 1989; Verbrugge & Jette, 1994), starting with limitations in mobility, then evolving to limitations in instrumental activities of daily living (IADL) and finally, evolving to limitations in basic activities of daily living (ADL).

However, recent research considers physical disability as a dynamic, reversible process in which an individual can move between different types and severities of disability and recovery (partial or complete) (Guralnik & Ferrucci, 2002). However, existent evidence on these topics is mostly based on developed countries like the U.S. (Gill & Kurland, 2003), Portugal (Nogueira & Reis, 2014) or South Korea (Lee & Park, 2006), documenting the experience of specific groups such as the institutionalized (Gill, Robinson, & Tinetti, 1997; Hardy & Gill, 2004), those with only IADL limitations (Seidel, Jagger, Brayne, Matthews, & MRC FAS, 2009) or those undergoing recovery after hospitalization (Boyd et al., 2008).

Research in developing countries is limited and several papers report recovery as an indirect result of their main objectives. While calculating disability-free life expectancy, researchers report that Mexicans aged 65+ have a higher probability of recovering from an ADL limitation than Puerto Ricans (Payne, 2015); others find no gender differences in the probability of recovering from an ADL limitation in São Paulo, Brazil (Andrade, Guevara, Lebrão, de Oliveira-Duarte, & Ferreira-Santos, 2011); additional research finds that around 34% of a national sample of older Mexican Americans recovered within two years, with factors like fewer depressive symptoms, higher BMI, and younger age being associated with their recovery (Al Snih, Markides, Ostir, Ray, & Goodwin, 2003). Finally, in China, older women have a lower probability than older men of recovering after reporting an ADL limitation (Gu & Zeng, 2004).

Therefore, this paper seeks to fill this research gap by asking: How common is recovery from functional limitations among the elderly population of a developing country like Mexico? And, how do different socioeconomic and health-related conditions impact this recovery? The paper has two goals: to document the characteristics of older adults who, having reported a functional limitation, show some improvement in their level of physical disability compared to those who do not; and to examine the key determinants of the likelihood of recovery. We focus on the characteristics of individuals as well as their contexts of family network and dwelling. As the existing literature on recovery in a developing country is scarce, we must speak of it as a complete process that includes all dimensions of disability (ADLs, IADLs and mobility) instead of having specific situations like those previously mentioned. We achieve these goals by using data from Mexico of individuals aged 60 or older, with four waves of a longitudinal study that includes a national sample of older adults.

2. Data and Methods

2.1. Sample

Data come from the Mexican Health and Aging Study (MHAS), a nationally representative panel investigation of health and aging in Mexicans born in 1951 or earlier. The baseline data, consisting of 15,186 in-person interviews, were collected in 2001 (92% response rate) with follow-ups in 2003 (93% response rate), 2012 (88% response rate), and 2015 (90% response rate). In 2012, a new sample was added so that the 2001 and 2012 waves remain representative of older adults aged 50 or older who live in the community. Further information about the MHAS is available elsewhere (Wong, Michaels-Obregón, & Palloni, 2015). This study was approved by the Institutional Review Boards or Ethics Committees of the University of Texas Medical Branch in the United States, the Instituto Nacional de Estadística y Geografía (INEGI) and the Instituto Nacional de Salud Pública (INSP) in Mexico.

As baseline data was collected in 2001 with follow-ups in 2003, 2012 and 2015, the 9-year gap between waves 2 and 3 presented some statistical challenges. We therefore decided to focus on two separate samples (2001–2003 and 2012–2015) to see how recovery changed in these two periods separated by a little over a decade.

We separated our analyses into two parts to measure whether respondents aged 60 or older recovered from a physical disability. For the 2001–2003 sample, the data begin with 7,171 respondents aged 60 or older who answered the interview themselves. Next, we excluded 1,474 respondents who had missing information in either 2001 or 2003 for at least one of the five ADL components (eating, bathing/showering, transferring in/out of bed, getting dressed, and using the toilet), one of the four IADL components (shopping, managing money, taking medications, and preparing meals) or one of the three mobility components (climb a flight of stairs, walk one-half mile, and lift heavy objects), as we are unable to accurately determine the individuals’ change in functionality between 2001 and 2003 (with mobility components being the most commonly omitted). Hence, our analytical sample consisted of 5,697 respondents for the descriptives and 4,297 for the regression, as some respondents had missing information in one or more covariates.

For 2012–2015, the sample consists of 11,254 respondents aged 60 or older in 2012 who answered the interview themselves. We exclude 1,335 respondents who had missing information in either 2012 or 2015 for at least one of the ADL, IADL, or mobility components (with components in mobility being the most commonly omitted) and also excluded the 2,711 respondents who were deceased by 2012. Thus, our analytical sample consisted of 7,208 respondents for the descriptives and 5,810 for the regression due to missing information in one or more covariates.

The main reason for missing data is because we included respondents aged 60 or older who had complete information on all of the individual activities (for ADL, IADL and mobility), as we believe this is one of our main contributions to the existing literature as we point out in section 1. Having complete information on all activities allows us to more accurately state if a respondent improved (partially or completely), remained the same or worsened in his/her condition.

2.2. Measures

The MHAS questionnaire asks the respondent if, because of a health issue, s/he has difficulty in performing several activities, excluding those difficulties believed to last less than three months. To obtain more accurate measurements, we combined this question with one asking whether the respondent needed help (from a person or from special equipment) to perform this activity. The end result is a dichotomous variable for each of the five ADL, four IADL and three mobility items listed above that combines the inability to perform and the need for help. Then, each respondent is categorized as having no physical limitations; having a physical limitation in one dimension (mobility only, ADL only or IADL only); having a physical limitation in two dimensions (ADL and mobility, IADL and mobility, or ADL and IADL); or having a limitation in all three dimensions. The full background for the creation of these variables can be found in (Díaz-Venegas, Reistetter, Wang, & Wong, 2016).

Each respondent was classified as recovered, worsened or the same. Specifically, “recovered” indicated that, by the 2-year follow-up in 2003 or the 3-year follow-up in 2015, s/he reported limitations in fewer physical dimensions. For example, a person with limitations in mobility only (one dimension) in 2001 will be classified as recovered if the same person reported no physical limitations in 2003. Similarly, a person with limitations in IADL and ADL (two dimensions) in 2012 would be classified as “worsened” if the same person reported limitations in three dimensions in 2015. It is worth mentioning that we consider “recovery” as including both partial and full recovery, as we believe disability is a reversible process and the elderly might regain some independent living that could improve their overall wellbeing.

2.3. Independent Variables

The following variables were constructed using data from 2001 and 2012. Age was measured as two dichotomous variables with respondents aged 60–74 (reference) and aged 75 or older. This division allowed us to examine the differences in the extent and covariates of recovery among the youngest old (60–74) and the oldest old (75+) in our samples. Rural residence was measured with a dichotomous variable as respondent living in a community with fewer than 100,000 inhabitants (= 1) or not (= 0). Marital status was measured as a dichotomous variable indicating if the respondent was married (= 1) or single, divorced, widowed or separated (= 0). Total cognition score (continuous, range 0–80) included tests measuring visuospatial ability (0–2 points), visuospatial memory (0–2 points), verbal learning (0–8 points), verbal recall (0–8 points) and visual scanning (0–60) in 2001. For 2012, the scores were the same except for visuospatial ability and memory (0–6 points each), giving a total cognition score of 88 points instead of 80. To adjust both years to the same 0–80 range, we followed the methodology of Michaels-Obregón, Mejía-Arango, and Wong (2014). Number of depressive symptoms was expressed as a dichotomous variable, to capture whether the respondent reported an elevated number of symptoms (CES-D score ≥ 5) using previous literature and validated for the MHAS (Aguilar-Navarro, Fuentes-Cantú, Ávila-Funes, & García-Mayo, 2007). Self-rated health (SRH) was scored to indicate poor SRH (= 1) and all other categories (excellent, very good, good, fair = 0). Previous literature (Viruell-Fuentes, Morenoff, Williams, & House, 2011) has identified that, when asked about health status in Spanish, the word “fair” (translated as “regular” in Spanish) carries a more positive connotation compared to the same question administered in English. Thus, we have included “fair” along with the other positively connotated words. The self-reported diagnosis of hypertension, diabetes, cancer, respiratory problems, heart condition or stroke was expressed as a dichotomous variable for each. Average individual monthly income was expressed as three separate dichotomous variables that included: indebted or no income (reference), earnings of less than 5,000 Mexican Pesos (MXP) and earnings of 5,000 Mexican Pesos (MXP) or more (1 USD is roughly equivalent to 22.50 MXP at current exchange rates). We used the imputed income values provided by the MHAS, which follows a similar method as the US Health and Retirement Study, based on a bracketed unfolding technique to reduce non-response and the method of sequence of regressions with a SAS-based software (IVEware). More details on the full process are available at Wong and Espinoza (2004). Health Insurance coverage measured as any type (= 1) vs. not covered (= 0). We consider whether the respondent received help from family and/or friends in a dichotomous variable measuring social support (= 1) vs. no support (= 0).

Finally, we included gender, measuring male (= 0) or female (= 1), and years of education, indicated by three dichotomous variables: respondents with zero years of formal education (reference), between 1 and 6 years of education, and with 7 or more years of education. These variables were taken at baseline in 2001 and in 2012.

2.4. Statistical Analysis

We presented our descriptive results by documenting the distribution of the sample according to the number of limited dimensions in 2001 and 2003. Then, we described the status in 2003, measuring whether the respondent had the same number of limited dimensions, if his/her condition worsened, if it improved, or if the respondent had died. Finally, we used a multinomial logistic regression to determine the risk of transitioning to each status in 2003 (worsened, improved or died) relative to remaining with the same number of limited dimensions (reference category), after controlling for the covariates listed above. The process was then repeated for the 2012–2015 sample. We use Stata 15.1 for the analysis (StataCorp, 2015).

3. Results

First, we mention some descriptive characteristics (data not shown) to differentiate the older adults in each period. In 2001, on average, respondents were 69.1 years old, had 3.5 years of education, 43.0% lived in rural areas, 54.3% were female, they had a total cognition score of 32.6 points, 62.7% had some type of health insurance, 42.4% reported being diagnosed with hypertension, 17.5% with diabetes, and the cohort had on average 3.8 symptoms of depression.

In 2012, on average, respondents were 71.1 years old, had 3.9 years of education, 53.9% lived in rural areas, 55.3% were female, they had a total cognition score of 35.1 points, 85.4% had some type of health insurance, 46.5% reported being diagnosed with hypertension, 23.2% reported having diabetes and the cohort had 3.6 symptoms of depression on average.

In both samples, missing cases were, on average, older, has fewer years of education, more were living in rural areas and fewer had insurance coverage. Further, the two baseline samples show significant differences in the percentage of those covered by any type of health insurance and those diagnosed with diabetes, which is one of the most prevalent chronic conditions in Mexico.

Table 1A shows the distribution of respondents stratified by the number of dimensions with limitations in 2001 and 2003. The diagonal of the table (highlighted in grey) contains the frequencies of those reporting the same number of limitations after 2 years. The percentages below the diagonal are those that improved, i.e., had a reduced number of limited dimensions, and those above the diagonal are the cases that worsened. The percentages showed that improvement or recovery played an important role in the 2-year status of physical function in this sample. Among the 1,238 cases who partially or fully recovered (below the diagonal, obtained by adding 710+122+216+32+74+84), more than half (710 cases) transitioned from one to zero limited dimensions; another 27% (338 cases, obtained by adding 122+216) transitioned from two to zero or one limited dimension. Further, 15% (188 cases) showed signs of recovery after being limited in all three dimensions, with 32 out of those 188 cases reporting zero limitations after two years.

Table 1A.

Distribution of Mexican adults aged 60 or older by the number of dimensions with physical limitations in 2001 and 2003

		2003					Total
		0	1	2	3	Died
2001 ***	0	1,905 66.9	646 22.7	143 5.0	55 1.9	97 3.4	2,846
	1	710 38.3	701 37.8	242 13.1	94 5.1	105 5.7	1,852
	2	122 20.3	216 35.9	119 19.8	88 14.6	57 9.4	602
	3	32 8.1	74 18.6	84 21.2	112 28.2	95 23.9	397

Note: Authors’ own elaboration with data from the Mexican Health and Aging Study (Mexican Health and Aging Study, 2001, 2003). Values shown are unweighted and row percentages are the bottom number. The diagonal is highlighted in grey to show those had the same number of limited dimensions after two years. We omitted 100 respondents who had missing information in 2001 and were deceased by 2003. Chi-square tests were calculated to establish whether the differences between the number of limited dimensions in 2001 and 2003 were significant.

Significance:

^***p ≤ .001.

Overall, a large proportion of cases that started with no (66.9%) or one (37.8%) type of limitation remained with the same number of limitations at follow-up. About one-half remained with the same number of limited dimensions two years later. The majority (92%) of those who remain with the same number of limited dimensions started with zero or one limitation in 2001. Of those who got worse, about half (51%) moved from zero to one type of limitation.

Table 1B shows the distribution of respondents stratified by the number of limited dimensions in 2012 and 2015. The diagonal of the table (highlighted in grey) contains the frequencies of those with the same number of limitations after 3 years. Among the 1,176 cases of those who partially or fully recovered (below the diagonal, obtained by adding 570+102+259+23+88+134), around 48% (570 cases) went from having limitations in one dimension to zero, and another 31% (361 cases, obtained by adding 102+259) transitioned from two to zero or one limited dimension. Further, around 21% (245 cases) showed signs of recovery after being limited in all three dimensions, with 23 out of 245 reporting zero limitations at three years.

Table 1B.

Distribution of Mexican adults aged 60 or older by the number of dimensions with physical limitations in 2012 and 2015

		2015					Total
		0	1	2	3	Died
2012 ***	0	1,895 59.2	845 26.4	221 6.9	92 2.9	146 4.6	3,199
	1	570 24.4	892 38.1	423 18.1	242 10.3	214 9.1	2,341
	2	102 10.4	259 26.4	250 25.5	224 22.8	146 14.9	981
	3	23 3.4	88 12.8	134 19.5	257 37.5	184 26.8	686

Note: Authors’ own elaboration with data from the Mexican Health and Aging Study (Mexican Health and Aging Study, 2012, 2015). Values shown are unweighted and row percentages are the bottom number. The diagonal is highlighted in grey to show those who remained with the same number of limited dimensions after three years. Chi-square tests were calculated to establish whether the differences between the number of limited dimensions in 2012 and 2015 were significant.

Significance:

^***p ≤ .001.

A large proportion of cases that started with no (59.2%) or one (38.1%) type of limitation also had the same number of limitations at follow-up. Overall, out of the 7,207 respondents, 45% had the same number of limited dimensions three years later, while 21% recovered, 28% got worse, and 9% died. Around 85% of those who remained with the same number of limited dimensions had zero or one limitation in 2012. Of those who got worse, 41% moved from zero to one type of limitation.

In summary, our results show evidence of similar patterns in 2-year (2001–2003) and 3-year (2012–2015) recovery, in the sense that respondents were able to partially or even fully recover in the two baseline samples. The overall percentage of recovery (whether partial or full) varied between the 2001–2003 sample (around 21%) and the 2012–2015 sample (around 18%).

Table 2A presents the characteristics of the older adults in 2001 according to their functional limitations status in 2003. More older respondents (aged 75+) had worsened condition compared to respondents aged 60–74 (28.1% vs. 21.2%). Further, those who retained the same number of limited dimensions had, on average, 3.9 years of educational attainment. This is higher than the average for respondents whose condition worsened, improved or who died (2.9, 2.6 and 2.6 years of educational attainment, respectively). This tendency was also seen in the total cognition score, with higher average scores for those in the “same” category (32.8 points) than for those in the other categories (25.7 points for those who worsened, 29.1 points for those who improved and 24.9 points for those who died).

Table 2A.

Distribution of Mexican adults aged 60 or older in 2001 by disability status based on the number of dimensions with physical limitations in 2003

Characteristics in 2001	Disability status in 2003
	Same	Worse	Improved	Died
Age 60–74 ***	53.1	21.0	21.5	4.4
Age 75+	38.7	26.6	22.4	12.3
Female ***	45.5	25.1	23.9	5.5
Male	54.9	18.9	19.1	7.1
Married ***	51.8	21.6	21.6	5.0
Not married	46.7	23.3	21.9	8.1
Rural ***	46.9	23.9	21.2	6.0
Urban	52.0	21.0	20.6	6.4
Years of education (avg.) ***	4.0	3.0	3.1	2.9
No education	43.1	24.1	24.7	8.1
Between 1 and 6 years	50.9	22.8	20.9	5.4
7 years or more	60.6	16.2	18.1	5.1
Cognition score (avg.) **	35.0	29.4	31.2	26.2
Poor self-reported health ***	36.1	22.2	29.9	11.8
Other self-reported health	53.4	22.3	19.6	4.7
No income ***	48.3	21.4	23.5	6.8
Less than 5,000 Mexican Pesos	48.4	23.7	21.5	6.4
Over 5,000 Mexican Pesos	58.9	16.9	20.0	4.2
With health insurance	51.4	21.8	20.8	6.0
Without health insurance	47.2	23.1	23.1	6.6
Receives social support	49.7	22.1	21.6	6.6
No social support	50.8	22.6	22.2	4.4
With hypertension ***	45.8	23.4	23.5	7.3
Without hypertension	52.5	21.6	20.4	5.5
With diabetes ***	42.0	25.7	22.6	9.7
Without diabetes	51.2	21.7	21.5	5.6
With cancer	40.2	26.5	23.5	9.8
Without cancer	49.7	22.4	21.7	6.2
With respiratory disease ***	40.2	21.6	27.8	10.4
Without respiratory disease	50.4	22.5	21.2	5.9
With heart condition ***	35.7	20.7	30.7	12.9
Without heart condition	50.2	22.5	21.3	6.0
With stroke ***	37.5	21.7	23.4	17.4
Without stroke	50.1	22.4	21.6	5.9
Depressive symptoms (avg.) ***	3.4	3.9	4.5	4.9
With depression	41.9	22.9	26.9	8.3
Without depression	57.7	21.8	16.4	4.1
Unweighted N	2,837	1,268	1,238	454

Note: Authors’ own elaboration with data from the Mexican Health and Aging Study (Mexican Health and Aging Study, 2001, 2003). Values shown are unweighted. All values are percentages unless noted and each row adds up to 100%. Chi-square tests were calculated to establish whether the differences between the categories of physical limitation and each covariate were significant.

Significance:

^*p ≤ .05;

^**p ≤ .01;

^***p ≤ .001.

Finally, poor self-rated health and chronic conditions showed an interesting story. On one side, compared to those without chronic diseases, 2-year mortality was higher among those who reported any of the six conditions included in the analysis, particularly for stroke (30.5% with stroke vs. 7.4% without), and around 21% of respondents improved despite having a chronic condition. However, for the cases reporting poor self-rated health, respiratory diseases, heart condition, and depressive symptoms, those with the condition at baseline were more likely to improve their physical functionality compared to those without the condition, a somewhat surprising result.

Table 2B presents the characteristics of the older adults in 2012 according to their functional limitations in 2015. Unlike in the previous table, respondents aged 75+ saw their condition worsen at a very similar level as respondents aged 60–74 (28.3% vs. 27.8%). Those who remained with the same number of limited dimensions had, on average, 4.3 years of educational attainment and an average cognition score of 37.4 points and, once again, these were higher than the average educational attainment and average total cognition scores for respondents whose condition worsened, improved or who died.

Table 2B.

Distribution of Mexican adults aged 60 or older in 2012 by disability status based on the number of dimensions with physical limitations in 2015

Characteristics in 2012	Disability status in 2015
	Same	Worse	Improved	Died
Age 60–74 ***	49.8	27.3	16.9	6.0
Age 75+	34.4	31.5	14.7	19.4
Female ***	42.4	31.8	17.2	8.6
Male	50.0	23.9	15.2	10.9
Married ***	47.7	26.7	17.2	8.4
Not married	42.3	31.3	14.9	11.5
Rural **	43.3	29.4	17.2	10.1
Urban	47.4	27.6	15.7	9.3
Years of education (avg.) ***	5.3	4.1	4.1	3.6
No education	37.4	31.6	18.7	12.3
Between 1 and 6 years	44.4	29.0	16.8	9.8
7 years or more	57.6	23.6	12.5	6.3
Cognition score (avg.) ***	39.4	35.1	35.8	28.2
Poor self-reported health ***	32.3	27.0	21.1	19.6
Other self-reported health	48.2	28.7	15.4	7.7
No income ***	43.4	30.6	17.6	8.4
Less than 5,000 Mexican Pesos	43.4	29.0	16.8	10.8
Over 5,000 Mexican Pesos	55.5	24.0	13.4	7.1
With health insurance	45.8	28.3	16.3	9.6
Without health insurance	45.4	29.4	16.0	9.2
Receives social support	45.5	28.5	16.3	9.7
No social support	47.1	27.3	16.4	9.2
With hypertension ***	42.1	30.4	16.9	10.6
Without hypertension	49.4	26.3	15.8	8.5
With diabetes ***	39.2	30.7	16.5	13.6
Without diabetes	48.0	27.5	16.3	8.2
With cancer ***	33.3	25.5	18.2	23.0
Without cancer	46.0	28.5	16.3	9.2
With respiratory disease ***	38.8	26.8	17.1	17.3
Without respiratory disease	46.3	28.5	16.2	9.0
With heart condition ***	35.8	23.4	21.2	19.6
Without heart condition	46.2	28.6	16.1	9.1
With stroke *	39.4	25.1	20.6	14.9
Without stroke	45.9	28.5	16.2	9.4
Depressive symptoms (avg.) ***	3.0	3.7	4.1	4.3
With depression	38.2	30.3	19.6	11.9
Without depression	51.6	27.2	13.7	7.5
Unweighted N	3,294	2,047	1,177	690

Note: Authors’ own elaboration with data from the Mexican Health and Aging Study (Mexican Health and Aging Study, 2012, 2015). Values shown are unweighted. All values are percentages unless noted and each row adds up to 100%. Chi-square tests were calculated to establish whether the differences between the categories of physical limitation and each covariate were significant.

Significance:

^*p ≤ .05;

^**p ≤ .01;

^***p ≤ .001.

As for chronic conditions, the 3-year mortality was higher among those who reported any of the six chronic conditions included in the analysis, especially for heart conditions (25.4% with stroke vs. 8.9% without) but, on average, 17% of the respondents improved despite having any chronic condition. This time, only respondents with cancer or a heart condition were more likely to improve their physical functionality compared to those without the condition.

A common finding in the two analyses is that those who reported a heart condition seemed to have a higher likelihood of recovering from physical limitations than those who reported other conditions.

Table 3A shows the results of the multinomial logistic regression for 2001–2003, expressed in relative-risk-ratios (RRR) and taking the same number of limited dimensions as the reference. First, for mortality, we found that the relative risk of dying was the highest for those with older age (75+) by a factor of 2.63 compared to those aged 60–74. The risk of getting worse (1.50) or improving (1.30) was also greater in those with older age, relative to maintaining the same number of limited dimensions. Further, compared to men, women had a higher relative risk of getting worse (1.48) or improving (1.47), but men were more likely to die by the 2-year follow-up, a result consistent with existing literature that suggests that women live longer than men but do so with more disability (Leveille, Resnick, & Balfour, 2000). Similarly, reporting poor self-rated health and having 5 or more symptoms of depression also increased the relative risk of reporting a change in the status of the respondent. The RRR of dying versus staying the same was 1.95 for poor self-rated health and 2.00 for depressive symptoms. The next highest RRR was for dying versus improving was 1.73 for poor self-rated health and 1.99 for depressive symptoms. The lowest RRR was for dying versus getting worse was 1.25 for poor self-rated health and 1.20 for depressive symptoms.

Table 3A.

Relative risk ratios of 2-year physical limitation status in Mexican adults aged 60 or older, 2001–2003

Variables at baseline (2001)	Status at 2-year follow-up
	Worse	Improved	Deceased
	(Reference: Same)
Demographic
Aged 75 or older	1.50***	1.37**	2.63***
Location size (ref.: 100,000 inhabitants or more)
Less than 100,000 inhabitants	0.99	0.97	0.70*
Gender (ref.: Men)
Women	1.48***	1.47***	0.68*
Education (ref.: No education)
Between 1 and 6 years	1.11	0.89	0.89
7 years or more	0.93	0.85	1.19
Marital Status (ref.: Non-married)
Married	1.06	1.15	0.80
Socioeconomic
Social support
Help from neighbors and/or children	0.94	0.98	1.45
Monthly income
(ref.: No or negative income)
Less than 5,000 Mexican Pesos	1.21	1.05	0.96
5,000 Mexican Pesos or more	0.84	0.90	0.55*
Insurance
Any coverage	0.90	0.88	0.88
Health and function variables
Poor self-rated health	1.25*	1.73***	1.95***
Depression (ref.: CES-D of 4 or more)	1.20*	1.99***	2.03***
Total cognition score	0.98***	0.99*	0.97***
Hypertension	1.11	1.06	1.28
Diabetes	1.36**	1.02	1.74**
Cancer	1.43	1.06	0.93
Respiratory diseases	1.10	1.53**	1.80*
Heart condition	1.60*	2.36***	2.19*
Stroke	0.89	1.05	1.58
Constant	0.41***	0.24***	0.11***
Unweighted N	4,297
Pseudo-R 2	0.05

Note: Authors’ own elaboration with data from the Mexican Health and Aging Study (Mexican Health and Aging Study, 2001, 2003). “Same” means the respondent had the same number of dimensions of physical ability with a limitation in 2001 and in 2003 and is the reference category. “Worse” means the respondent had more dimensions of physical limitation in 2003 than in 2001. “Improved” means the respondent had fewer dimensions of physical limitation in 2003 than in 2001. MXP = Mexican Pesos (1 USD is roughly equivalent to 22.50 MXP at the current exchange rates).

Significance:

^*p ≤ .05;

^**p ≤ .01;

^***p ≤ .001. Confidence intervals omitted due to space constraints. Values shown are unweighted.

Finally, in terms of the effect of chronic conditions, those reporting diabetes had a higher relative risk of dying (1.74) followed by getting worse (1.36), compared to remaining the same. In contrast, and while heart conditions (2.19) and respiratory diseases (1.80) increased the relative risk of dying compared to remaining the same, these conditions also increased the relative risk of improving (2.36 for heart conditions and 1.53 for respiratory illnesses). Overall, looking at the relative-risk-ratios significantly associated with these health conditions at baseline, the odds of improving were higher than the odds of getting worse. We performed sensitivity analyses of these results by excluding the deceased cases from the multivariate analysis, and obtained even stronger relative risk ratios; that is, when we considered only the cases that survived over the 2-year follow-up in our model of the dynamics of functionality, we overestimated the effects of significant variables such as age and chronic diseases on disability transitions.

Table 3B shows the results of the multinomial logistic regression for 2012–2015, expressed in RRR and taking the same number of limited dimensions as the reference. Similar to Table 3A, we found that the relative risk of dying was the highest in those with older age (75+) by a factor of 2.87 compared to age 60–74. Women had a higher relative risk of getting worse (1.48) or improving (1.22), but men were more likely to die by the 3-year follow-up. Reporting poor self-rated health or showing signs of depression also increased the relative risk of reporting a change in the status of the respondent, particularly shifting towards an improvement in the physical conditions (RRR of 1.53 for self-rated health and 1.70 for symptoms of depression).

Table 3B.

Relative risk ratios of 3-year physical limitation status in Mexican adults aged 60 or older, 2012–2015

Variables in 2012	Status at 3-year follow-up
	Worse	Improved	Deceased
	(Reference: Same)
Demographic
Aged 75 or older	1.64***	1.29*	2.87***
Location size (ref.: 100,000 inhabitants or more)
Less than 100,000 inhabitants	1.07	1.13	0.90
Gender (ref.: Men)
Women	1.48***	1.22*	0.70**
Education (ref.: No education)
Between 1 and 6 years	0.96	0.83	1.09
7 years or more	0.80	0.59***	1.21
Marital Status (ref.: Non-married)
Married	0.98	1.20*	0.98
Socioeconomic
Social support
Help from neighbors and/or children	1.05	1.03	1.06
Monthly income
(Ref.: No or negative income)
Less than 5,000 Mexican Pesos	0.97	0.99	0.90
5,000 Mexican Pesos or more	0.80*	0.89	0.75
Insurance
Any coverage	0.97	1.06	1.22
Health and function variables
Poor self-rated health	1.00	1.53***	2.29***
Depression (ref.: CES-D of 4 or more)	1.33***	1.70***	1.33*
Total cognition score	0.99**	1.00	0.96***
Hypertension	1.23**	1.08	1.10
Diabetes	1.33***	1.08	2.01***
Cancer	1.20	1.66*	2.75***
Respiratory diseases	0.99	1.04	2.03***
Heart condition	1.08	1.37	2.57***
Stroke	1.03	1.27	0.49
Constant	0.47***	0.21***	0.25***
Unweighted N	5,808
Pseudo-R 2	0.05

Note: Authors’ own elaboration with data from the Mexican Health and Aging Study (Mexican Health and Aging Study, 2012, 2015). “Same” means the respondent had the same number of dimensions of physical limitation in 2012 and in 2015 and is the reference category. “Worse” means the respondent had more dimensions of physical limitation in 2015 than in 2012. “Improved” means the respondent had fewer dimensions of physical limitation in 2015 than in 2012. MXP = Mexican Pesos (1 USD is roughly equivalent to 22.50 MXP at the current exchange rates).

Significance:

^*p ≤ .05;

^**p ≤ .01;

^***p ≤ .001. Confidence intervals omitted due to space constraints. Values shown are unweighted.

Finally, for the effect of chronic conditions, all respondents had a higher relative risk of dying for all chronic conditions, except hypertension (respondents had a higher RRR of worsening their condition) and stroke (none of the RRRs were statistically significant). This time, the effect we saw for the 2001–2003 period regarding improvement in respondents’ physical condition despite reporting one or more chronic conditions was not present. Only those respondents who reported having cancer had a slightly higher RRR of improving from their physical limitations (1.66), but this was overshadowed by the RRR of dying (2.75).

Additionally, we conducted an ancillary analysis (not shown) by adding one more category to the multinomial logistic regression, basically splitting recovery into two categories. Depending on the number of limited dimensions of physical ability, respondents could worsen, recover partially, recover fully, stay the same or die. Overall, our results remained mostly unchanged when the alternative definition of recovery was explored.

To gauge the joint overall effect of the covariates on the 2-year and 3-year transitions in the number of limited functionality areas, Tables 4A and 4B show the predicted probability of a respondent falling into each of the four statuses in 2003 and in 2015, for different profiles of individuals in 2001 and 2012, stratified by gender. Because we considered an exhaustive group of change categories, the row percentages added to 100%. Although we present all the results, we focus here on the recovery state.

Table 4A.

Predicted probabilities of 2-year status for selected cases of Mexican adults aged 60 or older in 2001 by gender, 2001–2003

Case	Status at 2-year follow-up
		Same	Worse	Improved	Deceased
Case 1: Covariates at means	Men	59.3	19.0	17.4	4.4
	Women	48.5	24.8	23.2	3.5
Case 2: Age (60–74) at baseline (2001)	Men	60.0	18.0	17.4	4.6
	Women	49.7	23.6	23.0	3.7
Case 3: Age (75+) at baseline (2001)	Men	49.7	21.9	19.0	9.4
	Women	40.0	28.0	24.6	7.4
Case 4: Age (75+), depression, poor self-rated health, and diabetes	Men	29.7	22.5	26.7	21.1
	Women	24.8	28.8	31.6	14.8
Case 5: Age (75+), depression, poor self-rated health, and no diabetes	Men	35.0	19.9	31.5	13.6
	Women	28.9	25.0	36.8	9.3

Note: Authors’ own elaboration with data from the Mexican Health and Aging Study (Mexican Health and Aging Study, 2001, 2003). Covariates included age, gender, years of education, urban/rural, marital status, social support, monthly income, health insurance, poor self-rated health, and presence of depression, total cognition score and six chronic conditions (hypertension, diabetes, cancer, respiratory problems, heart conditions and stroke). In all cases, the covariates not indicated are at their means. Values shown are unweighted.

Table 4B.

Predicted probabilities of 3-year status for selected cases of Mexican adults aged 60 or older in 2012 by gender, 2012–2015

Case	Status at 3-year follow-up
		Same	Worse	Improved	Deceased
Case 1: Covariates at means	Men	53.8	23.8	16.0	6.4
	Women	44.7	33.2	17.6	4.5
Case 2: Age (60–74) in 2012	Men	56.7	22.3	15.8	5.2
	Women	47.2	31.4	17.6	3.8
Case 3: Age (75+) in 2012	Men	44.1	28.4	15.9	11.6
	Women	35.7	39.0	17.1	8.2
Case 4: Age (75+), depression, poor self-rated health, and diabetes	Men	25.7	25.0	20.2	29.1
	Women	22.8	35.5	21.5	20.2
Case 5: Age (75+), depression, poor self-rated health, and no diabetes	Men	33.1	24.1	24.2	18.6
	Women	28.6	33.5	25.2	12.7

Note: Authors’ own elaboration with data from the Mexican Health and Aging Study (Mexican Health and Aging Study, 2012, 2015). Covariates included age, gender, years of education, urban/rural, marital status, social support, monthly income, health insurance, poor self-rated health, presence of depression, total cognition score and six chronic conditions (hypertension, diabetes, cancer, respiratory problems, heart conditions and stroke). In all cases, the covariates not indicated are at their means. Values shown are unweighted.

In Table 4A, the first case analyzes gender differences, and keeps all other covariates at their means. On average, women had a higher probability (23.2%) of 2-year recovery than men (17.4%) and a slightly lower chance of dying (3.5% vs. 4.4%). Cases 2 and 3 highlight younger vs. older adults, setting age at 60–74 (Case 2) and at 75+ (Case 3) while keeping the other covariates at their means. For both genders, younger age was associated with an estimated higher probability of remaining with the same number of limited dimensions. Older age was associated with a higher risk of death (9.4% for age 75+ versus 4.6% for age 60–74 for men, and 7.4% for age 75+ versus 3.7% for age 60–74 for women); similar age associations were evident for the probability of a worse status at the 2-year follow-up. However, one peculiar pattern emerges in the age effect for recovery, as shown by the regression results; holding all else at mean values, older adults are similarly or slightly more likely to improve compared to their younger counterparts (19.0% versus 17.4% for men, and 24.6% versus 23.0% for women).

Next, Cases 4 and 5 illustrate the effect of chronic diseases, highlighting an important chronic disease for the Mexican population: diabetes. In Case 4, we fixed the baseline age to an older person (aged 75+), with depression, with poor SRH and with diabetes, with all other variables at their means. Case 5 was similar, except without diabetes. The presence of the disease was associated with a higher probability of transitioning to a worse status and to dying, and with a lower likelihood of staying in the same status. The probability of improving was higher for cases without diabetes at baseline compared to those with diabetes (36.8% versus 31.6% for women).

In Table 4B, we performed the same analysis but focused on the 2012–2015 period. Case 1 analyzes gender differences, and keeps all other covariates at their means; on average, women had a slightly higher probability (17.6%) of 3-year recovery than men (16.0%) and a lower chance of dying (4.5% vs. 6.4%). In Cases 2 and 3, for both genders, younger age was associated with an estimated higher probability of remaining with the same number of limited dimensions. Older age was associated with a higher risk of death (in men, 11.6% for age 75+ versus 5.2% for age 60–74; in women, 8.2% for age 75+ versus 3.8% for age 60–74); similar age associations were evident for the probability of a worse status at the 3-year follow-up. In contrast to Table 4A, the age effect for recovery was not significantly different as the probabilities of recovery for men and women were quite similar. In Case 4, we once again fixed age to that of an older person (aged 75+), with depression, with poor SRH and with diabetes, setting all other variables at their means. Case 5 was similar, except without diabetes. The presence of the disease was associated with a slightly higher probability of transitioning to a worse status for women. For men, the probability of worsening or improving was nearly identical and also showed a lower likelihood of staying in the same status and a higher likelihood of improving if the respondent did not have diabetes.

4. Discussion and Conclusions

Using a national sample of older Mexican adults aged 60+, we documented 2-year (2001–2003) and 3-year (2012–2015) transitions in physical function among older adults, focusing on three dimensions: mobility, ADL and IADL limitations, and a construct that captured the dynamics of physical function: staying with the same number of limitations, increasing the number (worsening), decreasing the number (improving) or dying.

Our results confirmed several expectations. First, the majority of cases with one limited dimension presented with mobility limitations, which is expected given that mobility tends to be the first function that is lost among older adults, followed by IADL and ADL (Jette & Branch, 1981). Second, improvement in physical limitations seemed quite prevalent, being at least as prevalent as transitioning to a worse state. Third, there was heterogeneity in the patterns of change in functionality. The relative risks of changing status varied by age, by gender and by baseline health-related variables such as poor SRH, being depressed and having a chronic condition like diabetes, respiratory illnesses, and/or heart conditions.

Our results raise some questions as well. In the 2-year transitions (2001–2003), respondents who reported any of the chronic conditions at baseline had a higher relative risk of either getting worse or dying, but surprisingly, also of improving rather than remaining with the same number of limited dimensions. In fact, for those with conditions like heart disease at baseline, the RRR was higher for improving versus remaining the same compared to any other status.

These findings suggest that respondents with initial poor SRH or depression might be seeking and/or receiving personal care or some treatment to recover from their condition (Callahan et al., 2005), which would promote recovery from physical limitations at follow-up as well. Similarly, we speculate that those with heart conditions or respiratory illnesses might be inclined to seek and/or receive social support (Janevic et al., 2004), as well as be forced to alter their lifestyle to include more physical activity, better nutrition, and reducing or eliminating alcohol and tobacco (Nelson et al., 2007), thus improving their quality of life and promoting recovery from physical limitations. The results in the 3-year transitions (2012–2015) also showed some of these effects, but age acted as a mitigating factor that reduced the impact of the chronic conditions among the elderly.

Other authors have noted the result of having a more-favorable trajectory in one health dimension while the initial state is worse in another. Bishop and colleagues (2016) interpret such results as suggestive of floor and ceiling effects, whereas poor initial health in one domain seems less susceptible to declining even further. This line of thought can be further examined to verify the presence of floor or ceiling effects in our research.

Our work made several other contributions to the extant literature. Our approach, by including 2001–2003 and 2012–2015, seems to confirm observed patterns of recovery and common determinants across time periods. We also drew attention to the importance of simultaneously including all dimensions of physical function such as mobility, IADL and ADL for older adults, instead of analyzing each separately or as an index. Similarly, our results support the point that this line of work on functionality, while focusing on recovery or non-recovery, should also include death as a possible outcome (Peek, Patel, & Ottenbacher, 2005), in order to consider selective attrition in panel studies.

Nevertheless, our research also has the limitation of the lack of a clear definition of recovery from a physical limitation, as the existing literature normally centers on a specific situation (e.g., recovering after being hospitalized), a specific disease (e.g., recovering from a stroke) or a specific population (e.g., those institutionalized) using different methodologies and a different set of physical abilities (e.g., only IADL or only ADL). These discrepancies prevent the direct comparison of our results with any other reported data.

Physical limitations often involve a combination of social, psychological and physical factors that interact to complicate the process of measuring disability and recovery (Brandt Jr. & Pope, 1997). Older adults in Mexico are living longer and, as a result, are spending more time with a physical limitation and this, combined with the rising prevalence of chronic conditions like diabetes, has expanded the proportion of their lives spent with disability (Payne & Wong, 2019). However, older adults, particularly those living in rural areas, might show signs of high resilience by overcoming socioeconomic obstacles and being able to have a decent quality of life despite the presence of one or more chronic conditions and/or physical limitations (MacLeod, Musich, Hawkins, Alsgaard, & Wicker, 2016). Additional research is needed to further examine the trends of recovery among the elderly in other developing countries.

We close by emphasizing the commonality and salience of improvement or recovery as a specific outcome in studies of functionality, as most studies of physical limitations highlight the determinants and pace of deterioration. We also emphasize the importance of considering comorbidities — chronic physical diseases and other cognitive or mental health dimensions that could critically affect the trajectory of physical function. The effect of these conditions may at times have positive byproducts as we have speculated, and medical treatment for these diseases should continue to emphasize behaviors and life styles conducive to a parallel improvement in physical function.