**1. Introduction**

84 Rehabilitation Medicine

Sullivan KJ, Brown DA, Klassen T, Mulroy S, Ge T, Azen SP & Winstein CJ. (2007). Effects of

Tyson SF, Hanley M, Chillala J, Selley AB, Tallis RC. (2008). Sensory loss in hospital-

with function. *Neurorehabilitation and Neural Repair* Mar-Apr;22(2):166-72 Verheyden G, Nieuwboer A, De Wit L, Thijs V, Dobbelaere J, Devos H, Severijns D,

Wade DT & Hewer RL. (1987). Functional abilities after stroke: measurement, natural

Winstein CJ, Rose DK, Tan SM, Lewthwaite R, Chui HC & Azen SP. (2004). A randomized

*Repair* Oct;23(8):819-24

Apr;22(2):173-9

Feb;50(2):177-82

*and Rehabilitation* Apr;85(4):620-8

task-specific locomotor and strength training in adults who were ambulatory after stroke: results of the STEPS randomized clinical trial. *Physical Therapy* Dec;87(12):1580-602van Nes IJ, van Kessel ME, Schils F, Fasotti L, Geurts AC & Kwakkel G. (2009). Is visuospatial hemineglect longitudinally associated with postural imbalance in the postacute phase of stroke? *Neurorehabilitation and Neural* 

admitted people with stroke: characteristics, associated factors, and relationship

Vanbeveren S, De Weerdt W. (2008). Time course of trunk, arm, leg, and functional recovery after ischemic stroke. *Neurorehabilitation and Neural Repair* Mar-

history and prognosis. *Journal of Neurology, Neurosurgery and Psychiatry*

controlled comparison of upper-extremity rehabilitation strategies in acute stroke: A pilot study of immediate and long-term outcomes. *Archives of Physical Medicine* 

> As the older population grows dramatically around the world, it is important that health care providers be able to maintain people with an extended life expectancy in an active stage for as long as possible. Being independent in gross mobility functioning is an indicator of healthy and successful aging (Guralnik and Kaplan, 1989). An effective tool that is easy to use for identifying those at early stage of physical function decline is imperative for achieving this goal.

> As people age, a majority of elderly individuals develop physical disability. Such development follows a hierarchical order, starting from mobility, then spreading into instrumental activities of daily living (IADL), and finally ending in basic activities of daily living (BADL) (Pinsky et al., 1987; Barberger-Gateau et al., 1995; Barberger-Gateau et al., 2000). As disability in mobility occurs at an earlier stage of the disablement process, it may be an effective indicator by which to identify older adults in an early stage of physical function decline. Identifying such older adults is imperative in order to provide timely health promotion or early intervention programs.

> In the literature, mobility disability has been defined as at least one item requiring help, or being unable to perform independently using two items (climbing stairs and walking on a level surface) (Guralnik et al., 1994; Guralnik et al., 1995; Ostir et al., 1998) or three items (heavy housework, climbing stairs, and walking on a level surface) (Jette and Branch, 1981; Guralnik et al., 1994; Barberger-Gateau et al., 1995; Guralnik et al., 1995; Merrill et al., 1997; Barberger-Gateau et al., 2000; Ble et al., 2005; Yogev-Seligmann et al., 2008) in Rosow's scale (Rosow and Breslau, 1966). Using either two or three items, the disadvantage of a dichotomous mobility disability status is the inability to identify those with an intermediate status (those are becoming disabled, but are not yet disabled and thus require timely intervention).

<sup>\*</sup> Corresponding Author

The Hierarchical Status of Mobility Disability

**2.1 Measurements** 

**2.1.1 Four-level hierarchy of mobility disability** 

items disabled" and "3 items disabled", respectively.

been published elsewhere (Hsu, 2005; Chen et al., 2010).

Predicts Future IADL Disability: A Longitudinal Study on Ageing in Taiwan 87

the second stage, blocks in the selected administrative units were defined as clusters; and in the third stage, two respondents were selected systematically from the register in each block. An elderly cohort of 4,049 individuals aged 60 or older was first interviewed in 1989 and re-interviewed every three to four years until 2007 (response rate = 88.9 - 91.8%). An additional cohort of 2642 individuals aged 50 to 66 was recruited in 1996 and re-interviewed every four years until 2007 (response rate = 81.2 - 92.1%). Full details of the survey have

A total of 4,440 respondents were interviewed in the survey year 1999. For the purposes of our study, we first excluded those who were younger than 60 years old or who did not live in the community, leaving a sample size of 3,465. Another 4 and 47 respondents respectively had incomplete baseline data in mobility and other variables, such as education attainment, work status, spouse status, health lifestyle, disease status, self-rated health, BADL, and IADL, and were also excluded, further reducing the sample size to 3,414. In order to determine the risk of developing future IADL disability of each mobility disability group, we further excluded 977 individuals who already had IADL disability at baseline, leaving 2,437. Furthermore, we excluded 25 individuals due to loss of contact for follow-up and 285 individuals due to death before contributing any follow-up data on IADL in the years of 2003 and 2007. As a result, 2,127 community-dwelling older adults who had complete baseline data, were free of IADL disability at baseline, and contributed follow-up data at least once remained for further survival analysis of median age onset and the median

survival time of IADL disability for each hierarchical status of mobility disability.

For Cox regression, we used the extended-model approach for covariate adjustment: Model 1 = without adjustment; Model 2 = variables in Model 1 + demographics (age, sex, education, work, and spouse status); Model 3 = variables in Models 1 & 2 + health behaviors (smoking, alcohol, and exercise); Model 4 = variables in Models 1 & 2 & 3 + health status (number of co-morbidities, self-rated health, the Center for Epidemiologic Studies Depression (CESD), and Short Portable Mental Status Questionnaire (SPMSQ)). Data on 2,127 (at eight-year follow-up) and 2,073 (at four-year follow-up, with an additional 54 samples excluded due to loss of contact for follow-up in 2003) individuals were entered into Model 1. The amount of data entered into Model 4, however, decreased dramatically because participants younger than 65 were not interviewed for their cognitive function in the survey. In total, data on cognitive function (SPMSQ) were missing for 574 (eight-year follow-up) and 557 (four-year follow-up) individuals, and data on the CESD scale were missing for 8 individuals. As a result, we excluded the data from 582 (eight-year follow-up) and 565 (four-year follow-up) individuals when running Model Four in Cox regression.

This study extracted from the survey questionnaire three variables assessing gross mobility: heavy housework, climbing up 2 to 3 floors, and walking 200 to 300 m. Those reporting no difficulty in performing all three items were in the "mobility able" group, whereas those reporting difficulty performing only one item were categorized as "1 item disabled". Those reporting difficulty in any two of the three items or in all three items were treated as "2

The use of a summed number of tasks labeled as difficult has been proposed as a way to categorize the severity of BADL disability (Hing and Bloom, 1991; Manton et al., 1993). The item-wised hierarchical structure of mobility disability has been investigated only by Wang and colleagues using two items (Wang et al., 2005) in a cross-sectional study. They reported that these hierarchies could identify participants with different physical health and performance levels. Thus the item-wise hierarchy of mobility disability is able to monitor the status of mobility disability and to identify those at the stage of decline.

By using item-wise definitions of mobility hierarchy, previous studies in Taiwan have reported that individuals with more advanced mobility disability are associated with more concurrent dependence in IADL and BADL (Chen et al., 2010; Yeh et al., 2010). We therefore predicted a longitudinal relationship between hierarchical status of mobility disability and IADL disability. However, a literature review reveals that the item-wise hierarchical status of mobility disability for identifying individuals at higher risk of further IADL decline, which often follows the development of mobility disability (Pinsky et al., 1987; Barberger-Gateau et al., 1995; Barberger-Gateau et al., 2000), has not been substantiated. Furthermore, the median age of onset and the required time for 50% of people at different levels of mobility disability to develop IADL disability have not been reported. The required length of time to develop future IADL disability after the onset of each hierarchical mobility disability status is worthwhile to ascertain so that health care providers will be able to estimate how much time they have for early interventions.

Besides the current mobility disability status that might predict future disablement, other risk factors have been reported in the literature, including age, sex, spouse status (Reynolds and Silverstein, 2003), educational level, current working status, cigarette smoking, alcohol consumption, exercise habits (Miller et al., 2000; Sarkisian et al., 2000), number of comorbidities (Reynolds and Silverstein, 2003), self-rated health (Cornette, 2005), depressive symptoms (Sarkisian et al., 2000; Kazama et al., 2010), and cognition (Reynolds and Silverstein, 2003; Cornette, 2005; Yochim et al., 2008). These will be used, in this study, as covariates to ascertain the significance of mobility disability status in predicting future IADL disablement.

Thus, this study aimed to investigate the predictive validity of a four-level item-wise hierarchical mobility disability status for future IADL disability, using longitudinal data from a national representative sample. The specific purposes were (1) to ascertain the longitudinal relationship between hierarchies of mobility disability and IADL using the hazard ratio of the hierarchical mobility disability status in developing IADL disability across four and eight years of follow up. In order to ascertain the significant contribution of the hierarchical mobility status to IADL disability, we adjusted the potential risk factors that have been reported in the literature; and (2) to report the median age onset and the survival time for 50% of individuals to development of IADL disability (median survival time) in each hierarchical mobility disability stage.

#### **2. Methods**

Data were obtained from the 1999, 2003, and 2007 "Survey of Health and Living Status of the Elderly in Taiwan," a population-based longitudinal study with a nationally representative random sample. The sampling was conducted with a three-stage equal probability method. In the first stage, samples were stratified into administrative units; in the second stage, blocks in the selected administrative units were defined as clusters; and in the third stage, two respondents were selected systematically from the register in each block. An elderly cohort of 4,049 individuals aged 60 or older was first interviewed in 1989 and re-interviewed every three to four years until 2007 (response rate = 88.9 - 91.8%). An additional cohort of 2642 individuals aged 50 to 66 was recruited in 1996 and re-interviewed every four years until 2007 (response rate = 81.2 - 92.1%). Full details of the survey have been published elsewhere (Hsu, 2005; Chen et al., 2010).

A total of 4,440 respondents were interviewed in the survey year 1999. For the purposes of our study, we first excluded those who were younger than 60 years old or who did not live in the community, leaving a sample size of 3,465. Another 4 and 47 respondents respectively had incomplete baseline data in mobility and other variables, such as education attainment, work status, spouse status, health lifestyle, disease status, self-rated health, BADL, and IADL, and were also excluded, further reducing the sample size to 3,414. In order to determine the risk of developing future IADL disability of each mobility disability group, we further excluded 977 individuals who already had IADL disability at baseline, leaving 2,437. Furthermore, we excluded 25 individuals due to loss of contact for follow-up and 285 individuals due to death before contributing any follow-up data on IADL in the years of 2003 and 2007. As a result, 2,127 community-dwelling older adults who had complete baseline data, were free of IADL disability at baseline, and contributed follow-up data at least once remained for further survival analysis of median age onset and the median survival time of IADL disability for each hierarchical status of mobility disability.

For Cox regression, we used the extended-model approach for covariate adjustment: Model 1 = without adjustment; Model 2 = variables in Model 1 + demographics (age, sex, education, work, and spouse status); Model 3 = variables in Models 1 & 2 + health behaviors (smoking, alcohol, and exercise); Model 4 = variables in Models 1 & 2 & 3 + health status (number of co-morbidities, self-rated health, the Center for Epidemiologic Studies Depression (CESD), and Short Portable Mental Status Questionnaire (SPMSQ)). Data on 2,127 (at eight-year follow-up) and 2,073 (at four-year follow-up, with an additional 54 samples excluded due to loss of contact for follow-up in 2003) individuals were entered into Model 1. The amount of data entered into Model 4, however, decreased dramatically because participants younger than 65 were not interviewed for their cognitive function in the survey. In total, data on cognitive function (SPMSQ) were missing for 574 (eight-year follow-up) and 557 (four-year follow-up) individuals, and data on the CESD scale were missing for 8 individuals. As a result, we excluded the data from 582 (eight-year follow-up) and 565 (four-year follow-up) individuals when running Model Four in Cox regression.

#### **2.1 Measurements**

86 Rehabilitation Medicine

The use of a summed number of tasks labeled as difficult has been proposed as a way to categorize the severity of BADL disability (Hing and Bloom, 1991; Manton et al., 1993). The item-wised hierarchical structure of mobility disability has been investigated only by Wang and colleagues using two items (Wang et al., 2005) in a cross-sectional study. They reported that these hierarchies could identify participants with different physical health and performance levels. Thus the item-wise hierarchy of mobility disability is able to monitor

By using item-wise definitions of mobility hierarchy, previous studies in Taiwan have reported that individuals with more advanced mobility disability are associated with more concurrent dependence in IADL and BADL (Chen et al., 2010; Yeh et al., 2010). We therefore predicted a longitudinal relationship between hierarchical status of mobility disability and IADL disability. However, a literature review reveals that the item-wise hierarchical status of mobility disability for identifying individuals at higher risk of further IADL decline, which often follows the development of mobility disability (Pinsky et al., 1987; Barberger-Gateau et al., 1995; Barberger-Gateau et al., 2000), has not been substantiated. Furthermore, the median age of onset and the required time for 50% of people at different levels of mobility disability to develop IADL disability have not been reported. The required length of time to develop future IADL disability after the onset of each hierarchical mobility disability status is worthwhile to ascertain so that health care providers will be able to

Besides the current mobility disability status that might predict future disablement, other risk factors have been reported in the literature, including age, sex, spouse status (Reynolds and Silverstein, 2003), educational level, current working status, cigarette smoking, alcohol consumption, exercise habits (Miller et al., 2000; Sarkisian et al., 2000), number of comorbidities (Reynolds and Silverstein, 2003), self-rated health (Cornette, 2005), depressive symptoms (Sarkisian et al., 2000; Kazama et al., 2010), and cognition (Reynolds and Silverstein, 2003; Cornette, 2005; Yochim et al., 2008). These will be used, in this study, as covariates to ascertain the significance of mobility disability status in predicting future IADL

Thus, this study aimed to investigate the predictive validity of a four-level item-wise hierarchical mobility disability status for future IADL disability, using longitudinal data from a national representative sample. The specific purposes were (1) to ascertain the longitudinal relationship between hierarchies of mobility disability and IADL using the hazard ratio of the hierarchical mobility disability status in developing IADL disability across four and eight years of follow up. In order to ascertain the significant contribution of the hierarchical mobility status to IADL disability, we adjusted the potential risk factors that have been reported in the literature; and (2) to report the median age onset and the survival time for 50% of individuals to development of IADL disability (median survival time) in

Data were obtained from the 1999, 2003, and 2007 "Survey of Health and Living Status of the Elderly in Taiwan," a population-based longitudinal study with a nationally representative random sample. The sampling was conducted with a three-stage equal probability method. In the first stage, samples were stratified into administrative units; in

the status of mobility disability and to identify those at the stage of decline.

estimate how much time they have for early interventions.

each hierarchical mobility disability stage.

disablement.

**2. Methods** 

## **2.1.1 Four-level hierarchy of mobility disability**

This study extracted from the survey questionnaire three variables assessing gross mobility: heavy housework, climbing up 2 to 3 floors, and walking 200 to 300 m. Those reporting no difficulty in performing all three items were in the "mobility able" group, whereas those reporting difficulty performing only one item were categorized as "1 item disabled". Those reporting difficulty in any two of the three items or in all three items were treated as "2 items disabled" and "3 items disabled", respectively.

The Hierarchical Status of Mobility Disability

Sex (n, %) a, e

Educational level (n, %) <sup>a</sup>

Cigarette smoking (n, %) b, c

Alcohol consumption (n, %) <sup>a</sup>

Number of co-morbidities f

stroke, cancer, arthritis)

Self Perceived Health Status (n, %)a, d

Exercise (n, %)\*, <sup>c</sup>

f

Work status (missing data, n=7) (n, %)b, c

Spouse status (missing data, n=1) (n, %)b, c

(hypertension, DM, cardiac disease,

Cognition\_SPMSQ f 8.6 ± 0.8

Depression\_CESD-10 a 3.4 ± 4.3

no statistical significance between item 2 and item 3.

Predicts Future IADL Disability: A Longitudinal Study on Ageing in Taiwan 89

percentages of men; had higher educational levels, larger percentages currently working, and spouses; smoked, drank alcohol, and exercised; and had a lower number of co-morbidities, a

> Mobility Able (n=1531)

Age (yrs) ¥, a 70 (64, 73) 73 (67, 77) 72 (67.5, 75) 71 (68, 75)

 Men 997 (65.1%) 173 (49.2%) 58 (34.5%) 29 (42.0%) Women 534 (34.9%) 186 (51.8%) 110 (65.5%) 40 (58.0%)

Illiterate 332 (21.7%) 116 (32.3%) 68 (40.5%) 31 (44.9%) Elementary school 761 (49.6%) 163 (45.4%) 82 (48.8%) 31 (44.9%) Junior or senior high school 324 (21.2%) 61 (17.0%) 15 (8.9%) 6 (8.7%) Above college 114 (7.5%) 19 (5.3%) 3 (1.8%) 1 (1.5%)

No 1153 (75.3%) 320 (89.1%) 156 (92.9%) 61 (88.4%) Yes 378 (24.7%) 39 (10.9%) 12 (7.1%) 8 (11.6%)

No 410 (26.8%) 135 (37.6%) 61 (36.3%) 27 (39.1%) Yes 1121 (73.2%) 224 (62.4%) 107 (63.7%) 42 (60.9%)

No 1073 (70.1%) 289 (80.5%) 139 (82.7%) 56 (81.2%) Yes 458 (29.9%) 70 (19.5%) 29 (17.3%) 13 (18.8%)

No 1049 (68.5%) 290 (80.8%) 141 (83.9%) 61 (88.4%) Yes 482 (31.5%) 69 (19.2%) 27 (16.1%) 8 (11.6%)

No 539 (35.2%) 136 (37.9%) 75 (44.6%) 36 (52.2%) Yes 992 (64.8%) 223 (62.1%) 93 (55.4%) 33 (47.8%)

 Worse 227 (14.8%) 128 (35.6%) 79 (47.0%) 41 (59.4%) Healthier 1304 (85.2%) 231 (64.4%) 89 (53.0%) 28 (40.6%)

(n=1081)

(n=1504)

¥: median (q1,q3), a significant differences were found between "mobility able group and all other (1 item, 2 items, and 3 items) disabled" groups; b significant differences were found between "mobility able and 2 items disabled"; c significant differences were found between "mobility able and 3 items disabled" group; d significant differences was found between "1 item disabled and 3 items disabled groups"; e significant differences was found between "1 item disabled and 2 items disabled groups";

Table 1. Demographic and health-related information at baseline (year of 1999) (n=2127).

0.7 ± 0.9 1.1 ± 1.0 1.3 ± 1.1 1.5 ± 1.2

8.1 ± 1.3 (n=139)

7.5 ± 6.1 (n=166)

8.0 ± 1.3 (n=53)

8.6 ± 7.0 (n=64)

8.3 ± 1.1 (n=280)

5.9 ± 5.7 (n=341)

1 Item Disabled (n=359)

2 Items Disabled (n=168)

3 Items Disabled (n=69)

better perceived health status, better cognition, and lower depression symptom scores.
