Skip to main content
NCBI home page
The Journals of Gerontology Series B: Psychological Sciences and Social Sciences logoLink to The Journals of Gerontology Series B: Psychological Sciences and Social Sciences
. 2018 Nov 2;75(3):630–639. doi: 10.1093/geronb/gby133

Changes in Physical Functioning as Short-Term Predictors of Mortality

Theresa Andrasfay 1,✉
Editor: Deborah Carr
PMCID: PMC7768692  PMID: 30388248

Abstract

Objectives

Both performance-based and self-reported measures of physical functioning are predictors of mortality. There has been relatively little research examining whether their changes predict mortality. This study examines whether 5-year changes in performance-based and self-reported measures of functioning predict subsequent mortality.

Method

Data are from the 2006 wave of the Social Environment and Biomarkers of Aging Study, 2011 wave of the Taiwan Longitudinal Study of Aging, and mortality follow-up through 2015. Gompertz proportional hazard models predict mortality from changes in ability to complete performance-based tests and changes in performance-based and self-reported functioning.

Results

Incident inability to complete at least one performance-based test of functioning is associated with twice the risk of subsequent 4-year mortality. Conditional on the baseline measurement, a one standard deviation (SD) decline in grip strength is associated with a 61% increased risk of 4-year mortality; a one-SD decline in walking speed and a one-SD increase in self-reported limitations are both associated with around a 40% increased risk of 4-year mortality. Conditional on the most recent measurement of functioning, prior change is not significantly associated with subsequent mortality.

Discussion

Repeated measures of performance-based and self-reported functioning are valuable in that they provide an updated measurement of functioning.

Keywords: Death and dying, Functional health status, Longitudinal methods, Measurement

As populations around the world have experienced rapid aging, understanding factors associated with the health and survival of older adults has become an important line of research. Physical functioning is one dimension of health and quality of life for older adults. In contrast, loss of functioning is understood to be a step in the disablement process through which individuals become less functionally independent (Fried, Herdman, Kuhn, Rubin, & Turano, 1991; Verbrugge & Jette, 1994).

Physical functioning is typically assessed through either self-reported physical limitations or performance-based tests. Though they both measure physical functioning, they capture different aspects and each have advantages and disadvantages. Self-reports are easily collected in interview-based studies; they represent individuals’ perceptions of difficulty with various tasks, and they are relevant to quality of life (Cornman et al., 2011; Guralnik, Branch, Cummings, & Curb, 1989). Like global self-assessed health, self-reports of physical limitations allow an individual to report about typical functioning as opposed to functioning at that particular moment. However, self-reports can also reflect pain, mental health, personality, understanding of the question, and cultural norms about expectations for physical ability (Cornman et al., 2011; Guralnik et al., 1989; Kempen et al., 1996; Reiman & Manske, 2011). Methodologically, self-reports can be problematic because many respondents report no difficulties in any domain of functioning, resulting in floor effects (Brach, VanSwearingen, Newman, & Kriska, 2002; Reuben et al., 2004; Sherman & Reuben, 1998).

Performance-based measures have been proposed as an objective alternative to self-reports to more directly assess functioning. These usually require a participant to complete some simple physical task using a measurement instrument or while being timed. Performance-based tests can assess both the ability to do a task and, through the performance on the test, the difficulty in completing it (Guralnik et al., 1989). Compared to self-reports, performance-based tests should theoretically be more comparable across populations, less reliant on language or culture, and better able to detect variation among people who report no physical limitations (Guralnik et al., 1989; Reuben et al., 2004; Sherman & Reuben, 1998). However, recent studies have questioned the comparability of these measures across populations due to variability in study protocols, cultural norms, and lifestyle differences (Glei, Goldman, Ryff, & Weinstein, 2018; Jeune et al., 2006).

Performance-based measures of physical functioning have been shown to reflect specific aspects of health. Handgrip strength is associated with overall body strength (Rantanen et al., 2003), cognitive function (Praetorius Björk, Johansson, & Hassing, 2016), and cardiovascular health (Lee, Peng, Chiou, & Chen, 2016), whereas declines in gait speed are associated with cognitive functioning (Watson et al., 2010), hypertension (Rosano et al., 2011), and inflammation (Verghese et al., 2011). Poor physical performance on relatively simple tasks can thus signal deteriorating health. Disadvantages of performance-based measures include their cost to implement in surveys—they require specialized equipment and are more burdensome on the respondent. To a lesser extent, performance-based tests are also subject to floor effects because some individuals cannot complete the test. Unlike more holistic self-reports, performance-based measurements are situation dependent, capturing functioning or effort at that specific time rather than typical functioning.

A large literature has demonstrated that cross-sectional measures of both self-reported and performance-based measures are independent predictors of future mortality. Self-reported physical limitations are independently associated with mortality even when performance tests are considered (Goldman, Glei, Rosero-Bixby, Chiou, & Weinstein, 2014; Melzer, Lan, & Guralnik, 2003; Reuben et al., 2004). Worse performance on or inability to complete tests of grip strength (Cooper, Kuh, & Hardy, 2010; Goldman et al., 2014; Leong et al., 2015; Nofuji et al., 2016; Rantanen et al., 2003; Xue, Beamer, Chaves, Guralnik, & Fried, 2010), peak expiratory flow (PEF; Goldman et al., 2014; Smith et al., 2013; Vaz Fragoso, Gahbauer, Van Ness, Concato, & Gill, 2008), gait speed (Blain et al., 2010; Cooper et al., 2010; Goldman et al., 2014; Nofuji et al., 2016), and chair stand speed (Cooper et al., 2010; Goldman et al., 2014) have all been associated with increased risk of mortality.

One of the stated advantages of performance-based measures is that they are more likely than self-reports to detect real change over time rather than variability in reporting over time (Buchman, Wilson, Boyle, Bienias, & Bennett, 2007; Guralnik et al., 1989; Hirsch, Buzková, Robbins, Patel, & Newman, 2012). Despite this supposed advantage, relatively little research has examined the ability of changes in these measures to predict mortality.

A small number of studies have examined whether declines in physical performance measures are associated with mortality independent of the baseline level of performance. More rapid declines in grip strength (De Buyser et al., 2016; Hirsch et al., 2012; Xue et al., 2010), lung function (Baughman et al., 2012; Mannino & Davis, 2006; Ryan et al., 1999), gait speed (Buchman et al., 2007; Hardy, Perera, Roumani, Chandler, & Studenski, 2007; Sabia et al., 2014; White et al., 2013), and chair stand speed (De Buyser et al., 2016) have been associated with an increased mortality risk. Results regarding declines in self-reports of physical function have been mixed, though self-reported improvements in walking ability have been associated with lower mortality risk (De Buyser et al., 2016; Hardy et al., 2007; Latham, 2016).

Although these studies show that declines are significant predictors of mortality, they are typically limited to Western countries, have small sample size, usually consider changes in only one or two domains of functioning, and exclude people who are unable to complete performance-based tests. Inability to complete at least one of the typical performance-based tests is common in older adults and assessing whether this crude initial measurement of functioning predicts mortality will help researchers and clinicians understand the value of these tests in older populations.

Very few studies have examined whether change in functioning is a significant predictor of mortality when considered alongside the most recent level of functioning (De Buyser et al., 2016; Sabia et al., 2014). If repeated measurements of functioning provide insight into subclinical changes in health, we would expect individuals with the same current level of functioning to have different mortality risks depending on how they arrive at this level of functioning. If the value of repeated measurements lies mainly in updating the information, we would expect that how an individual arrives at his or her current level of functioning does not improve predictions of mortality.

Our study addresses these gaps by assessing the predictive value of changes in both performance-based and self-reported assessments of physical functioning. Unlike prior studies examining change over time in performance, we include individuals who were unable to complete performance-based measures and systematically assess whether inability to perform at least one performance-based test is a predictor of mortality. We also examine change in functioning alongside the most recent measurement and the baseline measurement to determine whether repeated measurements can detect subclinical changes or whether they primarily provide an updated measurement of functioning.

Method

Data

Data are from the 2006 wave of the Social Environment and Biomarkers of Aging Study (SEBAS) and the 2011 wave of the Taiwan Longitudinal Study of Aging (TLSA). The TLSA began in 1989 as a nationally representative sample of the population aged 60 years and older, designed to assess the health and well-being of the older Taiwanese population. Comprised of a randomly selected subsample of the 1999 TLSA respondents, SEBAS began in 2000 as a biosocial survey with both an in-home interview and a hospital examination (Cornman et al., 2016). A set of in-home health assessments was introduced to the study beginning with the 2006 SEBAS wave, and these assessments were repeated in the 2011 TLSA wave for the subset of SEBAS participants. Additional details regarding sample selection and attrition can be found elsewhere (Cornman et al., 2016).

The analytic sample is restricted to those who participated in both the 2006 SEBAS wave and 2011 TLSA wave. Of the 951 individuals who participated in both waves, 887 have valid information on sociodemographic information, self-reported limitations, and performance measures in both waves, and of these, 776 were able to complete all performance-based tests in both waves. The outcome measure, survival status as of December 31, 2015, was ascertained through linkage with the Death Certificate Report System of the Ministry of Health and Welfare.

Measures

Performance-based measures

Four performance-based measurements of physical functioning were included in the home interview portion of SEBAS: grip strength, PEF, timed walk, and chair stands. Respondents were coded as completely missing a measurement if they refused to participate, if they did not understand the instructions, or if there was an equipment failure. Respondents completely missing one or more measurements (N = 64) have been excluded from the analysis, as these types of missing values do not provide any meaningful information about physical functioning. In contrast, when respondents had no measurement for a particular task because they were unable to perform the test, they were coded as unable to complete a task and were retained in the analytic sample because this type of missing value is likely to indicate poor performance and has been previously associated with mortality (Goldman et al., 2014).

Grip strength (in kilograms) was measured using a North Coast Hydraulic Hand Dynamometer. Respondents performed three trials on each hand by squeezing the handle of the dynamometer as hard as possible; the maximum of these trials is used as the measurement of grip strength in the analysis. Respondents were classified as unable to perform the tests if they met the exclusion criteria (recent injury, surgery, or severe pain in the hand, wrist, or arm), the interviewer or participant felt it was unsafe, the respondent tried but could not complete, or the respondent had to stop because of pain or discomfort.

PEF (in liters/minute) was measured using a TruZone peak flow meter. Respondents were asked to inhale and blow as hard and fast as possible into the peak flow meter; the maximum of three trials is used as the measurement of PEF in the analysis. Respondents were coded as unable to perform the tests if they met the exclusion criteria (recent surgery on the chest or abdomen, recent heart attack or heart problem, recent detached retina or eye surgery, or recent hospitalization for respiratory or lung infection), the interviewer or participant felt it was unsafe, the respondent tried but could not complete, or the respondent had to stop because of pain or discomfort.

Walking speed (in meters/second) was measured by timing the respondent as he or she walked 3 m at normal speed. The faster of two trials is used as the measurement of walking speed in the analysis. Respondents were coded as unable to perform the walk if the interviewer or participant felt it was unsafe or if the respondent tried but could not complete the walk. Respondents could use canes or walking aids if they normally did so when walking.

Chair stand speed (in stands/second) was measured by timing the respondent as he or she repeatedly stood up from a chair. Respondents were asked to sit on an armless chair against the wall with their arms crossed then stand up and sit back down five times as quickly as possible. Respondents were coded as unable to perform the chair stands if they were in a wheelchair, the interviewer or participant felt it was unsafe, or if the respondent tried but could not complete the stands. Chair stand speed was adjusted for chair height with the method used in Cornman and colleagues (2011); details are available in Supplementary Material 1.

Self-reports of physical limitations

In both waves, respondents were asked about their level of difficulty performing nine tasks: standing continuously for 15 min, standing continuously for 2 h, squatting, raising both hands overhead, grasping or turning objects, lifting or carrying 11–12 kg, running 20–30 m, walking 200–300 m, and walking up two or three flights of stairs. For each of these activities, respondents reported no difficulty (0), some difficulty (1), great difficulty (2), or unable to do (3). We constructed a scale by summing all nine items for a potential range of 0–27. This scale has been suggested as a way to capture the variability in physical limitations in older populations (Long & Pavalko, 2004). To reduce skewness, in the analytic models we added 0.5, and took the logarithm.

Statistical Analysis

Descriptive statistics are weighted by the longitudinal weight for the 2006–2011 SEBAS sample to account for differential response and attrition by age, sex, location, and other characteristics. We further adjusted these weights for inclusion in the analytic sample; additional details on the adjustment are available in Supplementary Material 1.

We fit Gompertz proportional hazards models on the analytic sample with the adjusted sample weights. In the Gompertz model, the hazard of mortality is assumed to increase exponentially with age; it has been shown to be a good fit for all-cause human mortality, especially at older ages (Horiuchi and Coale 1982; Juckett and Rosenberg 1993). Preliminary analyses revealed that the Gompertz survival form closely approximated the survival curve fit to the data using the nonparametric Cox model. We found no evidence that the proportional hazards assumption was violated. The time clock begins at the respondent’s age at his or her 2011 interview. Individuals were observed until their death or were censored on December 31, 2015. (One respondent was censored on December 31, 2014, due to possible emigration in 2015.)

We fit four sets of models to predict mortality from changes in physical functioning. Models 1 and 2 assess whether inability to complete at least one of the four performance-based tests (any inability) predicts mortality. We use this summary measure of any inability rather than task-specific inability because there are too few respondents with task-specific inability in each wave. In Model 1, we include an indicator for any inability in the 2006 wave and subsequent change in this indicator. In Model 2, we include an indicator for any inability in the 2011 wave and prior change in this indicator. These models produce equivalent predictions, but it is useful to examine both to understand which time of measurement provides the most predictive power.

Models 3 and 4 assess whether more fine-grained measures of changes in performance-based measures and changes in self-reported limitations predict mortality. In Model 3, we include the 2006 baseline functioning and change from 2006 to 2011 to determine if change provides additional information beyond the 2006 baseline measurement. In Model 4, we include the 2011 endline functioning and change from 2006 to 2011 to determine if change can provide additional predictive information beyond the most recent measurement. Because we fit a linear model for the logged hazard of mortality, the effects in the original scale are nonlinear, such that a steeper decline in functioning can have a greater impact than a small decline or an improvement in functioning. In these models, we restrict the analytic sample to the participants who could complete all performance-based tests in both waves. We separately fit Models 3 and 4 for each of the four performance-based tests.

We first include controls for (a) sex, years of education, and urban residence in 2011 and then add (b) self-reported physical limitations at both waves in order to test if changes in performance-based measures provide additional information beyond what could be obtained from survey responses alone. For comparison, we fit Models 3 and 4 for self-reported physical limitations in the absence of any performance-based measures. A summary of these models can be found in Supplementary Table 1.

To ease comparability between the different measures, we have converted all of the measures and changes to z scores based on the weighted mean and standard deviation of each measure at the 2006 wave. Results with the unstandardized measures can be found in Supplementary Table 2. A one unit increase for a performance measure corresponds to better function, whereas a one unit increase in the physical limitations scale corresponds to worse function (more reports of limitations). All analyses were conducted in R version 3.3.1 using the flexsurv package (Jackson, 2016).

Results

Weighted descriptive statistics are presented in Table 1. The average age at the 2011 interview was approximately 70 years. After approximately 4 years of follow-up, 11.2% had died by December 31, 2015. The scale of self-reported physical limitations increased on average from 3 in 2006 to 4.5 in 2011. For all four performance-based measures, the average change between the two waves was negative; however, a nontrivial number of participants experienced no decline or improved between the two study waves. On average, participants had 1 kg weaker grip strength, 20 L/min lower PEF, 0.1 m/s slower walking speed, and the same chair stand speed in 2011 compared to 2006. Participants were 3 percentage points more likely to experience any inability in performance-based tests in 2011 than they were in 2006.

Table 1.

Weighted Descriptive Statistics

Measure Mean (SD) or %
Demographic
 Age at 2011 survey 70.1 (8.7)
 Female 48.5%
 Urban residence 46.6%
 Years of completed education (0–17) 6.6 (4.7)
 Died by Dec 31, 2015 11.2%
 Mean follow-up (years) 3.9 (0.7)
Measures of physical functioning 2006 2011 Change
2006–2011
 Self-reported physical limitations scale (0–27) 3.0 (5.1) 4.5 (6.5) 0.3 (1.1)
Performance-based tests
 Any inability to complete performance-based tests 6.7% 9.9%
Grip strength
 Unable to perform grip strength test 1.9% 1.6%
 Grip strength (kg) 28.9 (10.4) 27.7 (10.3) −1.0 (6.4)
PEF
 Unable to perform PEF test 1.1% 1.8%
 PEF (L/min) 342.0 (135.3) 321.5 (141.4) −20.2 (88.4)
Walking speed
 Unable to perform walking speed 2.1% 3.1%
 Walking speed (m/s) 0.9 (0.3) 0.8 (0.3) −0.1 (0.3)
Chair stand speed
 Unable to perform chair stands 5.7% 8.4%
 Chair stand speed (stands/s) 0.5 (0.2) 0.5 (0.2) −0.01 (0.2)
Open in a new tab

Note: Data are from the 2006 wave of the Social Environment and Biomarkers of Aging Study (SEBAS), 2011 wave of the Taiwan Longitudinal Study of Aging, and Ministry of Health and Welfare. Data are weighted using the longitudinal weights for the 2006–2011 SEBAS sample adjusted for inclusion in the analytic sample (N = 887). PEF = peak expiratory flow.

Results from the models predicting mortality from inability to complete at least one of the performance-based tests (Models 1 and 2) are summarized in Table 2. In the absence of self-reported physical limitations, any inability in 2006 and subsequent change in this measure are both associated with a near-doubling of subsequent mortality risk. However, when any inability in 2011 is considered alongside prior change in this measure, only inability in 2011 is a significant predictor of mortality. When self-reported physical limitations are included in the model, neither inability nor change in inability significantly predicts mortality.

Table 2.

Gompertz Models Predicting Mortality From Any Inability to Complete Performance-Based Tests

Hazard ratio (95% confidence interval)
Modela Any inability in 2006 Change in ability 2006–2011 Any inability in 2011 Includes self-reported physical limitations?
Model 1a 1.96* (1.03, 3.73) 2.06** (1.25, 3.39) No
Model 1b 1.43 (0.67, 3.04) 1.60+ (0.96, 2.69) Yes
Model 2a 1.05 (0.56, 1.96) 1.96* (1.03, 3.73) No
Model 2b 1.12 (0.58, 2.18) 1.43 (0.67, 3.04) Yes
Open in a new tab

Note: Data are from the 2006 wave of the Social Environment and Biomarkers of Aging Study (SEBAS), 2011 wave of the Taiwan Longitudinal Study of Aging, and Ministry of Health and Welfare. Any inability is defined as inability to complete at least one performance-based test.

aModel 1a (2a): Adjusted for gender, education, and urban/rural residence and weighted by the longitudinal SEBAS weights adjusted for inclusion in analytic sample (N = 887). Model 1b (2b): Model 1a (2a) + the scale of self-reported limitations in 2006 and 2011.

+p< .10. *p < .05. **p < .01. ***p < .001.

Figure 1 displays the results from the models without controls for physical limitations (Models 1a and 2a). The average predicted probability of dying within 5 years of the 2011 interview is shown for each of four scenarios: (a) no inability in either wave, (b) no inability in 2006 and inability in at least one task in 2011, (c) inability in at least one task in 2006 and no inability in 2011, and (d) inability in at least one task in both waves. From this figure, it is clear that incident inability or sustained inability is associated with a substantially increased probability of subsequent mortality compared to those that were able to do all tests in both waves. Those with inability in at least one task in 2006 but no inability in 2011 had no increased probability of mortality compared to those that were able to do all tests in both waves.

Figure 1.

Figure 1.

Open in a new tab

Average predicted probability of dying within 5 years of the 2011 interview by inability to complete performance-based measures.

Results from the models predicting mortality from measures of physical functioning and changes in physical functioning (Models 3 and 4) are presented in Table 3. In the models predicting mortality from the initial 2006 measurement, 2006–2011 change, and basic sociodemographic controls (Model 3a), the baseline measurements of all four performance-based measures are associated with subsequent 4-year mortality; subsequent changes in both grip strength and walking speed are associated with subsequent 4-year mortality. For example, a one-SD higher baseline grip strength is associated with a 45% reduction in 4-year mortality (hazard ratio [HR] = 0.55) and a one-SD improvement relative to baseline grip strength is associated with a 38% reduction in 4-year mortality (HR = 0.62).

Table 3.

Gompertz Models Predicting Mortality From Functioning and Changes in Functioning

Hazard ratio (95% confidence interval)
Modela 2006 Change in functioning 2011 Includes self-reported limitations?
Functioning 2006–2011 Functioning
Grip strength
 Model 3a 0.55** (0.36, 0.84) 0.62*** (0.49, 0.79) No
 Model 3b 0.59* (0.38, 0.91) 0.66*** (0.51, 0.84) Yes
 Model 4a 0.89 (0.69, 1.16) 0.56** (0.37, 0.84) No
 Model 4b 0.91 (0.70, 1.19) 0.59* (0.38, 0.91) Yes
Peak flow
 Model 3a 0.59** (0.43, 0.81) 0.85 (0.66, 1.10) No
 Model 3b 0.63** (0.46, 0.87) 0.90 (0.70, 1.17) Yes
 Model 4a 1.20 (0.94, 1.53) 0.58** (0.42, 0.80) No
 Model 4b 1.23 (0.96, 1.57) 0.61** (0.44, 0.86) Yes
Walking speed
 Model 3a 0.74+ (0.53, 1.03) 0.70* (0.52, 0.94) No
 Model 3b 0.81 (0.56, 1.16) 0.78 (0.57, 1.06) Yes
 Model 4a 0.93 (0.71, 1.21) 0.73+ (0.52, 1.03) No
 Model 4b 0.95 (0.73, 1.25) 0.80 (0.55, 1.17) Yes
Chair stand speed
 Model 3a 0.72+ (0.52, 1.01) 0.92 (0.69, 1.23) No
 Model 3b 0.74 (0.52, 1.07) 0.97 (0.73, 1.28) Yes
 Model 4a 1.27 (0.92, 1.74) 0.71+ (0.51, 1.01) No
 Model 4b 1.30 (0.93, 1.80) 0.73 (0.50, 1.07) Yes
Self-reported limitations
 Model 3a 1.21 (0.88, 1.65) 1.40** (1.09, 1.79) Yes
 Model 4a 1.19 (0.94, 1.50) 1.23 (0.87, 1.73) Yes
Open in a new tab

Note: Data are from the 2006 wave of the Social Environment and Biomarkers of Aging Study (SEBAS), 2011 wave of the Taiwan Longitudinal Study of Aging, and Ministry of Health and Welfare. Hazard ratios refer to the effect of a one SD increase relative to the 2006 distribution.

aModel 3a (4a): Adjusted for gender, education, and urban/rural residence and weighted by longitudinal SEBAS weights adjusted for inclusion in analytic sample (N = 776). Model 3b (4b): Model 3a (4a) + the scale of self-reported limitations in 2006 and 2011.

+p < .10. *p < .05. **p < .01. ***p < .001.

In the models predicting mortality from the 2011 measurement, 2006–2011 change, and basic sociodemographic controls (Model 4a), only the most recent measure of functioning is a significant predictor of mortality; inclusion of the prior change does not significantly improve the prediction.

When self-reported limitations are included in these models (Model 3b and Model 4b), only the results for grip strength and PEF remain significant. The models with self-reported physical limitations alone are similar to the results with the performance-based measures in that change measured prospectively significantly predicts mortality but change measured retrospectively does not.

For ease of interpretation, Figure 2 displays the results from the models without controls for physical limitations (Models 3a and 4a) in terms of the predicted 5-year probability of mortality. The top panel displays the predicted 5-year probability of mortality for individuals who have average functioning in 2006 and between 2006 and 2011 experience (a) no change in functioning, (b) a one-SD decline in functioning, and (c) a one-SD improvement in functioning. Compared to those who maintained average grip strength, those who declined one-SD between 2006 and 2011 have on average a 5 percentage point increase in the probability of subsequent 5-year mortality, whereas those who improved one-SD have a 4 percentage point decline in the probability of subsequent 5-year mortality. Compared to those who maintained average walking speed, those who declined one-SD between 2006 and 2011 have on average a 3 percentage point increase in the probability of subsequent 5-year mortality, whereas those who improved one-SD have a 3 percentage point decline in the probability of subsequent 5-year mortality. Though not significant at the 5% level, subsequent declines and improvements in self-reported physical limitations exhibit similar associations with mortality.

Figure 2.

Figure 2.

Open in a new tab

Average predicted probability of dying within 5 years of the 2011 interview by functioning and change in functioning.

The lower panel displays the predicted 5-year probability of mortality from the models with retrospective change in functioning. The scenarios compare individuals who have average functioning in 2011 but arrived at this level of functioning either through (a) no change in functioning, (b) a one-SD decline in functioning, or (c) a one-SD improvement in functioning. There is no significant difference in the probability of subsequent mortality between these three scenarios; information about how individuals arrived at their current level of functioning does not improve mortality predictions.

Supplemental Analyses

In supplemental analyses (not shown), we considered alternative specifications of change in physical functioning between the two waves. We conducted the same analysis on a sample constructed through five sets of imputations to address item nonresponse from 64 respondents, distinct from inability to complete the tasks. We repeated the analysis using a Cox proportional hazards model, a model with fewer assumptions. The conclusions were the same with these specifications. In addition, we considered interactions between functioning and sex, but these were never significant in preliminary analyses. Finally, we considered models that incorporate both changes in ability and changes in level in the same model. These included modeling change in five categories (steep decline, moderate decline, no change, moderate improvement, and strong improvement) and constructing paths of functioning (always high, always low, high to low, low to high). All of these specifications result in the same broad conclusions: When measured prospectively, declines, including incident inability to complete a performance-based test, predict significantly increased risk of mortality. However, results regarding improvements in functioning were mixed and improvement was not consistently different than no change in functioning. Prior change is not a significant predictor of mortality if the most recent measure of functioning is included in the same model.

Discussion

Using the 2006 wave of SEBAS, 2011 wave of TLSA, and linked mortality data, we examined whether 5-year changes in simple measures of physical functioning predicted short-term mortality among a population of older Taiwanese adults.

First, we analyzed whether inability to complete performance-based tests of physical functioning in either wave predicts mortality. Previous research has not systematically considered changes in ability to complete these performance-based tasks, though inability to complete these tasks is increasingly common with age. We find that becoming unable to complete at least one performance-based task appears to represent a very severe decline in physical functioning and underlying health, as reflected in the substantial increase in mortality risk.

Our primary goal was to determine whether changes in the levels of physical functioning predict mortality among those who were able to complete the performance-based tests in both waves. Most studies that have examined declines in physical functioning have taken this approach. Consistent with prior literature, we find that conditional on baseline functioning, those with steeper declines in functioning typically have higher mortality in the subsequent 4 years.

We also analyzed changes in the levels of functioning in relation to the most recent measurement to determine whether prior change provided additional predictive information. The benefits of repeated measurements appear to derive from updating the measurement, rather than knowing a respondent’s history of functioning. This finding is consistent with previous studies of changes in grip strength and chair stand speed (De Buyser et al., 2016; Sabia et al., 2014). Thus, for prognostic purposes, there is little additional benefit to including prior change in functioning if the most recent measurements are available. For clinicians, these findings imply that patients’ current level of functioning is predictive of their future outcomes, regardless of how they reached this level of functioning.

Finally, for comparison, we included changes in a scale constructed from self-reports of physical limitations. One of the stated advantages of performance-based over self-reported measures of functioning is their ability to detect true change as opposed to reporting variability. When we included this scale in models predicting mortality from inability to complete at least one performance-based task, we found that inability to complete performance-based tasks was no longer statistically significant. It is likely that individuals who are unable to complete these tasks recognize the severity of their physical limitations and report accordingly. When we included this scale in the models with changes in performance-based measures, we found that only grip strength and PEF remained statistically significant in the presence of self-reports. The physical limitations scale asks several questions about mobility, and walking speed and chair stand speed may not provide additional information beyond what respondents already reported. Even though the respondents reported on their difficulty grasping objects, the grip strength test may capture additional variation in underlying strength. Respondents did not report on any aspect of lung function, so it is plausible that PEF provides additional information beyond the self-reports. These findings suggest that performance-based tests can capture variation in certain domains of functioning that is not identified by self-reports.

Like all studies, this study is affected by limitations that must be acknowledged. Due to the small number of respondents that were unable to complete each task, we could not examine whether inability to complete individual performance-based tasks predicts mortality. The number with inability to complete these tasks is likely small because individuals who were unable to complete a task in 2006 had an increased risk of death before the 2011 wave (Goldman et al., 2014). Given that our results suggest that inability to complete a performance-based test is associated with a substantial increase in mortality risk, it is likely that inability reflects extremely poor functioning of a separate class than the worst performers on the test. Researchers should examine whether changes in ability to complete these tasks predicts mortality in samples with more individuals unable to complete performance-based tests of functioning.

Though the performance measures have all been shown to have high test–retest reliability in controlled conditions, there is still potential for measurement error in the home interview setting, especially for walking speed and chair stand speed, which rely on an interviewer’s timing rather than an instrument (Jette, Jette, Ng, Plotkin, & Bach, 1999; Wolinsky, Miller, Andresen, Malmstrom, & Miller, 2005). As such, measurement error could bias the results toward finding no effect of the change in performance.

With only two waves of data, we are unable to examine long-term trajectories of physical functioning. Studies that have assessed trajectories over several periods have found that it is not uncommon for individuals with a long-term decline in a measure to experience transient improvements at intermediate waves, and we may be observing these transient improvements rather than a true increase in functioning (Xue et al., 2010). More than two waves of functioning would allow us to smooth over these aberrations. As more waves of longitudinal studies become available, future research should examine whether change measured over several time periods predicts mortality.

We are also unable to include those that died before the 2011 survey. Participants of the 2006 wave that died before the 2011 wave were older, less educated, more likely to be male, and had worse performance-based and self-reported physical functioning than those that participated in the 2011 wave. As a result, the analytic sample used in this study is positively selected in terms of initial health and thus we are not able to observe the full spectrum of changes in physical functioning that occur in this age range. This could limit the generalizability of our findings to individuals with relatively high initial functioning, though this limitation is common to all longitudinal studies.

Moreover, although these data are prospective, they are observational and thus we cannot make causal claims about whether interventions to preserve or improve physical functioning would reduce the risk of death in the near future. In particular, there may be unobserved covariates that predict both death and physical functioning. Future research should examine whether interventions to improve or maintain functioning would reduce mortality risk.

Supplementary Material

gby133_suppl_Supplementary_File_1
Click here for additional data file. (14KB, docx)
gby133_suppl_Supplementary_File_2
Click here for additional data file. (77.5KB, pdf)
gby133_suppl_Supplementary_File_3
Click here for additional data file. (17.2KB, docx)

Acknowledgments

I thank Noreen Goldman, Dana Glei, Germán Rodríguez, Jennifer Ailshire, and the anonymous reviewers for their comments and suggestions.

Funding

This work was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health (P2CHD047879, T32HD007163). The content is solely the responsibility of the author and does not necessarily represent the official views of the National Institutes of Health. Funding for the Social Environment and Biomarkers of Aging Study was provided by the Behavioral and Social Research Division of the U.S. National Institute on Aging (R01 AG16790, R01 AG16661).

Conflict of Interest

None reported.

References

  1. Baughman P., Marott J. L., Lange P., Martin C. J., Shankar A., Petsonk E. L., & Hnizdo E (2012). Combined effect of lung function level and decline increases morbidity and mortality risks. European Journal of Epidemiology, 27, 933–943. doi:10.1007/s10654-012-9750-2 [DOI] [PubMed] [Google Scholar]
  2. Blain H., Carriere I., Sourial N., Berard C., Favier F., Colvez A., & Bergman H (2010). Balance and walking speed predict subsequent 8-year mortality independently of current and intermediate events in well-functioning women aged 75 years and older. The Journal of Nutrition, Health & Aging, 14, 595–600. doi:10.1007/s12603-010-0111-0 [DOI] [PubMed] [Google Scholar]
  3. Brach J. S., VanSwearingen J. M., Newman A. B., & Kriska A. M (2002). Identifying early decline of physical function in community-dwelling older women: Performance-based and self-report measures. Physical Therapy, 82, 320–328. doi:10.1093/ptj/82.4.320 [PubMed] [Google Scholar]
  4. Buchman A. S., Wilson R. S., Boyle P. A., Bienias J. L., & Bennett D. A (2007). Change in motor function and risk of mortality in older persons. Journal of the American Geriatrics Society, 55, 11–19. doi:10.1111/j.1532-5415.2006.01032.x [DOI] [PubMed] [Google Scholar]
  5. Cooper R., Kuh D., & Hardy R; Mortality Review Group; FALCon and HALCyon Study Teams (2010). Objectively measured physical capability levels and mortality: Systematic review and meta-analysis. BMJ (Clinical Research Ed.), 341, c4467. doi:10.1136/bmj.c4467 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Cornman J. C., Glei D. A., Goldman N., Chang M. C., Lin H. S., Chuang Y. L.,…Weinstein M (2016). Cohort profile: The Social Environment and Biomarkers of Aging Study (SEBAS) in Taiwan. International Journal of Epidemiology, 45, 54–63. doi:10.1093/ije/dyu179 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Cornman J. C., Glei D., Rodríguez G., Goldman N., Hurng B. S., & Weinstein M (2011). Demographic and socioeconomic status differences in perceptions of difficulty with mobility in late life. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 66, 237–248. doi:10.1093/geronb/gbq087 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. De Buyser S., Petrovic M., Taes Y., Toye K., Kaufman J. M., Goemaere S., & Lapauw B (2016). Three year functional changes and long-term mortality hazard in community-dwelling older men. European Journal of Internal Medicine, 35, 66–72. doi:10.1016/j.ejim.2016.06.006 [DOI] [PubMed] [Google Scholar]
  9. Fried L. P., Herdman S. J., Kuhn K. E., Rubin G., & Turano K (1991). Preclinical disability: Hypotheses about the bottom of the iceberg. Journal of Aging and Health, 3, 285–300. doi:10.1177/089826439100300210 [Google Scholar]
  10. Glei D. A., Goldman N., Ryff C. D., & Weinstein M (2018). Physical function in US older adults compared with other populations: A multinational study. Journal of Aging and Health. doi:10.1177/0898264318759378\ [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Goldman N., Glei D. A., Rosero-Bixby L., Chiou S. T., & Weinstein M (2014). Performance-based measures of physical function as mortality predictors: Incremental value beyond self-reports. Demographic Research, 30, 227–252. doi:10.4054/DemRes.2013.30.7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Guralnik J. M., Branch L. G., Cummings S. R., & Curb J. D (1989). Physical performance measures in aging research. Journal of Gerontology, 44, M141–M146. doi:10.1093/geronj/44.5.M141 [DOI] [PubMed] [Google Scholar]
  13. Hardy S. E., Perera S., Roumani Y. F., Chandler J. M., & Studenski S. A (2007). Improvement in usual gait speed predicts better survival in older adults. Journal of the American Geriatrics Society, 55, 1727–1734. doi:10.1111/j.1532-5415.2007.01413.x [DOI] [PubMed] [Google Scholar]
  14. Hirsch C. H., Buzková P., Robbins J. A., Patel K. V., & Newman A. B (2012). Predicting late-life disability and death by the rate of decline in physical performance measures. Age and Ageing, 41, 155–161. doi:10.1093/ageing/afr151 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Horiuchi S., & Coale A. J (1982). A simple equation for estimating the expectation of life at old ages. Population Studies, 36, 317–326. doi:10.1080/00324728.1982.10409034 [DOI] [PubMed] [Google Scholar]
  16. Jackson C. H. (2016). flexsurv: A platform for parametric survival modelling in R. Journal of Statistical Software, 70, 1–33. doi:10.18637/jss.v070.i08 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Jette A. M., Jette D. U., Ng J., Plotkin D. J., & Bach M. A (1999). Are performance-based measures sufficiently reliable for use in multicenter trials? Musculoskeletal Impairment (MSI) Study Group. The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences, 54, M3–M6. doi:10.1093/gerona/54.1.M3 [DOI] [PubMed] [Google Scholar]
  18. Jeune B., Skytthe A., Cournil A., Greco V., Gampe J., Berardelli M.,…Robine J. M (2006). Handgrip strength among nonagenarians and centenarians in three European regions. The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences, 61, 707–712. doi:10.1093/gerona/61.7.707 [DOI] [PubMed] [Google Scholar]
  19. Juckett D. A., & Rosenberg B (1993). Comparison of the Gompertz and Weibull functions as descriptors for human mortality distributions and their intersections. Mechanisms of Ageing and Development, 69, 1–31. doi:10.1016/0047-6374(93)90068-3 [DOI] [PubMed] [Google Scholar]
  20. Kempen G. I., van Heuvelen M. J., van den Brink R. H., Kooijman A. C., Klein M., Houx P. J., & Ormel J (1996). Factors affecting contrasting results between self-reported and performance-based levels of physical limitation. Age and Ageing, 25, 458–464. doi:10.1093/ageing/25.6.458 [DOI] [PubMed] [Google Scholar]
  21. Latham K. (2016). “Taking a turn for the better”: Does self-reported recovery from walking limitation improve disability and mortality outcomes?The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 73, 1278–1291. doi: 10.1093/geronb/gbw113 [DOI] [PubMed] [Google Scholar]
  22. Lee W.-J., Peng L.-N., Chiou S.-T., & Chen L.-K (2016). Relative handgrip strength is a simple indicator of cardiometabolic risk among middle-aged and older people: A nationwide population-based study in Taiwan. PLoS One, 11, e0160876. doi:10.1371/journal.pone.0160876 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Leong D. P., Teo K. K., Rangarajan S., Lopez-Jaramillo P., Avezum A., Orlandini A.,…Kelishadi R (2015). Prognostic value of grip strength: Findings from the Prospective Urban Rural Epidemiology (PURE) study. The Lancet, 386, 266–273. doi:10.1016/S0140-6736(14)62000-6 [DOI] [PubMed] [Google Scholar]
  24. Long J. S., & Pavalko E (2004). Comparing alternative measures of functional limitation. Medical Care, 42, 19–27. doi:10.1097/01.mlr.0000102293.37107.c5 [DOI] [PubMed] [Google Scholar]
  25. Mannino D. M., & Davis K. J (2006). Lung function decline and outcomes in an elderly population. Thorax, 61, 472–477. doi:10.1136/thx.2005.052449 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Melzer D., Lan T. Y., & Guralnik J. M (2003). The predictive validity for mortality of the index of mobility-related limitation–results from the EPESE study. Age and Ageing, 32, 619–625. doi:10.1093/ageing/afg107 [DOI] [PubMed] [Google Scholar]
  27. Nofuji Y., Shinkai S., Taniguchi Y., Amano H., Nishi M., Murayama H.,…Suzuki T (2016). Associations of walking speed, grip strength, and standing balance with total and cause-specific mortality in a general population of Japanese elders. Journal of the American Medical Directors Association, 17, 184.e1–184.e7. doi:10.1016/j.jamda.2015.11.003 [DOI] [PubMed] [Google Scholar]
  28. Praetorius Björk M., Johansson B., & Hassing L. B (2016). I forgot when I lost my grip-strong associations between cognition and grip strength in level of performance and change across time in relation to impending death. Neurobiology of Aging, 38, 68–72. doi:10.1016/j.neurobiolaging.2015.11.010 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Rantanen T., Volpato S., Ferrucci L., Heikkinen E., Fried L. P., & Guralnik J. M (2003). Handgrip strength and cause-specific and total mortality in older disabled women: Exploring the mechanism. Journal of the American Geriatrics Society, 51, 636–641. doi:10.1034/j.1600-0579.2003.00207.x [DOI] [PubMed] [Google Scholar]
  30. Reiman M. P., & Manske R. C (2011). The assessment of function: How is it measured? A clinical perspective. The Journal of Manual & Manipulative Therapy, 19, 91–99. doi:10.1179/106698111X12973307659546 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Reuben D. B., Seeman T. E., Keeler E., Hayes R. P., Bowman L., Sewall A.,…Guralnik J. M (2004). Refining the categorization of physical functional status: The added value of combining self-reported and performance-based measures. The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences, 59, 1056–1061. doi:10.1093/gerona/59.10.M1056 [DOI] [PubMed] [Google Scholar]
  32. Rosano C., Longstreth W. T. Jr, Boudreau R., Taylor C. A., Du Y., Kuller L. H., & Newman A. B (2011). High blood pressure accelerates gait slowing in well-functioning older adults over 18-years of follow-up. Journal of the American Geriatrics Society, 59, 390–397. doi:10.1111/j.1532-5415.2010.03282.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Ryan G., Knuiman M. W., Divitini M. L., James A., Musk A. W., & Bartholomew H. C (1999). Decline in lung function and mortality: The Busselton Health Study. Journal of Epidemiology and Community Health, 53, 230–234. doi:10.1136/jech.53.4.230 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Sabia S., Dumurgier J., Tavernier B., Head J., Tzourio C., & Elbaz A (2014). Change in fast walking speed preceding death: Results from a prospective longitudinal cohort study. The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences, 69, 354–362. doi:10.1093/gerona/glt114 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Sherman S. E., & Reuben D (1998). Measures of functional status in community-dwelling elders. Journal of General Internal Medicine, 13, 817–823. doi:10.1046/j.1525-1497.1998.00245.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Smith M., Zhou M., Wang L., Peto R., Yang G., & Chen Z (2013). Peak flow as a predictor of cause-specific mortality in China: Results from a 15-year prospective study of ~170,000 men. International Journal of Epidemiology, 42, 803–815. doi:10.1093/ije/dyt079 [DOI] [PubMed] [Google Scholar]
  37. Vaz Fragoso C. A., Gahbauer E. A., Van Ness P. H., Concato J., & Gill T. M (2008). Peak expiratory flow as a predictor of subsequent disability and death in community-living older persons. Journal of the American Geriatrics Society, 56, 1014–1020. doi:10.1111/j.1532-5415.2008.01687.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Verbrugge L. M., & Jette A. M (1994). The disablement process. Social Science & Medicine, 38, 1–14. doi:10.1016/0277-9536(94)90294-1 [DOI] [PubMed] [Google Scholar]
  39. Verghese J., Holtzer R., Oh-Park M., Derby C. A., Lipton R. B., & Wang C (2011). Inflammatory markers and gait speed decline in older adults. The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences, 66, 1083–1089. doi:10.1093/gerona/glr099 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Watson N. L., Rosano C., Boudreau R. M., Simonsick E. M., Ferrucci L., Sutton-Tyrrell K.,…Newman A. B; Health ABC Study (2010). Executive function, memory, and gait speed decline in well-functioning older adults. The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences, 65, 1093–1100. doi:10.1093/gerona/glq111 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. White D. K., Neogi T., Nevitt M. C., Peloquin C. E., Zhu Y., Boudreau R. M.,…Zhang Y (2013). Trajectories of gait speed predict mortality in well-functioning older adults: The Health, Aging and Body Composition Study. The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences, 68, 456–464. doi:10.1093/gerona/gls197 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Wolinsky F. D., Miller D. K., Andresen E. M., Malmstrom T. K., & Miller J. P (2005). Reproducibility of physical performance and physiologic assessments. Journal of Aging and Health, 17, 111–124. doi:10.1177/0898264304272784 [DOI] [PubMed] [Google Scholar]
  43. Xue Q. L., Beamer B. A., Chaves P. H., Guralnik J. M., & Fried L. P (2010). Heterogeneity in rate of decline in grip, hip, and knee strength and the risk of all-cause mortality: The Women’s Health and Aging Study II. Journal of the American Geriatrics Society, 58, 2076–2084. doi:10.1111/j.1532-5415.2010.03154.x [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

gby133_suppl_Supplementary_File_1
Click here for additional data file. (14KB, docx)
gby133_suppl_Supplementary_File_2
Click here for additional data file. (77.5KB, pdf)
gby133_suppl_Supplementary_File_3
Click here for additional data file. (17.2KB, docx)

Articles from The Journals of Gerontology Series B: Psychological Sciences and Social Sciences are provided here courtesy of Oxford University Press

RESOURCES