Skip to main content

Table 1 Variable characteristics and model selection

From: Does being physically active prevent future disability in older people? Attenuated effects when taking time-dependent confounders into account

 

Baseline characteristics

Characteristic is missinga

Characteristic is time-varyingb

Characteristic's association withthe propensity for physical activity (treatment)c

p

Characteristic's association with transition to disability (outcome)d

p

Association with treatment and/or outcomee

Variables included in final modeling

Age at baseline (mean (SD))

74.1 (5.0)

0.0

NA

1.0 (1.0 - 1.0)

0.190

1.1 (1.0 - 1.1)

<0.001

outcome only

x

Gender (% female)

54.8

0.0

NA

1.2 (0.9 - 1.6)

0.360

0.9 (0.7 - 1.1)

0.250

none

 

Level of education (%)f

51.3

0.2

NA

2.0 (1.5 - 2.7)

<0.001

0.5 (0.4 - 0.6)

<0.001

both

x

Marital status (%)

 

22.4

6.6

 

0.016

 

0.032

both

x

 Single

4.9

  

reference

 

reference

   

 Married or partnership

63.1

  

2.1 (1.0 - 4.3)

0.055

0.7 (0.4 - 1.2)

0.220

  

 Widowed

27.3

  

1.4 (0.7 - 3.1)

0.340

1.0 (0.6 - 1.9)

0.960

  

 Separated

4.7

  

3.8 (1.4 - 10.3)

0.008

0.5 (0.2 - 1.3)

0.160

  

White matter pathology (%)f

 

0.0

NA

 

<0.001

 

<0.001

both

x

 Mild

44.4

  

reference

 

reference

   

 Moderate

30.9

  

0.9 (0.6 - 1.2)

0.420

1.6 (1.1 - 2.3)

0.009

  

 Severe

24.7

  

0.5 (0.3 - 0.7)

<0.001

3.2 (2.3 - 4.4)

<0.001

  

Stroke (%)g,f

7.2

0.0

NA

1.0 (0.5 - 1.8)

1.000

1.7 (1.1 - 2.6)

0.021

outcome only

x

Alcohol consumption (%)f

55.6

0.3

NA

1.1 (0.8 - 1.5)

0.670

0.9 (0.7 - 1.1)

0.320

none

 

Atrial fibrillation (%)g

7.5

23.0

5.3

1.1 (0.6 - 1.8)

0.820

1.7 (1.1 - 2.5)

0.013

outcome only

x

Angina pectoris (%)g

15.4

23.2

12.1

0.7 (0.5 - 1.0)

0.050

1.5 (1.0 - 2.1)

0.040

both

x

Anxiety or depressed mood (%)f,g

36.8

23.6

32.6

0.4 (0.3 - 0.5)

<0.001

1.8 (1.4 - 2.4)

<0.001

both

x

Body mass index (kg/m2; mean(SD))f

26.2 (4.2)

34.1

30.4

 

0.060

 

0.266

treatment only

 

 Middle tertile (range)

24.2 - 27.3

  

reference

 

reference

   

 First tertile (range)

16.4 - 24.2

  

1.2 (0.9 - 1.8)

0.250

1.3 (0.9 - 1.8)

0.140

  

 Last tertile (range)

27.3 - 47.8

  

0.8 (0.6 - 1.1)

0.150

1.3 (0.9 - 1.8)

0.190

  

Chronic pain (%)g

33.0

23.0

27.2

0.4 (0.3 - 0.5)

<0.001

1.2 (0.9 - 1.5)

0.300

treatment only

 

Complaints of gait disturbance (%)g

40.8

23.5

30.8

0.3 (0.2 - 0.4)

<0.001

2.6 (1.9 - 3.4)

<0.001

both

x

History of falls (%)g

28.6

22.7

39.6

0.6 (0.4 - 0.8)

<0.001

1.9 (1.4 - 2.5)

<0.001

both

x

Major depressive episode (%)f,g

6.1

22.5

9.1

0.6 (0.3 - 1.2)

0.130

2.9 (1.9 - 4.5)

<0.001

outcome only

x

MMSE (%)f

11.8

23.5

14.4

0.5 (0.3 - 0.7)

<0.001

4.1 (3.1 - 5.5)

<0.001

both

x

Osteoarthritis (%)g

28.1

23.0

22.7

0.7 (0.5 - 1.0)

0.032

0.8 (0.6 - 1.1)

0.170

treatment only

 

Peripheral vascular disease (%)g

7.2

24.1

6.6

0.5 (0.3 - 0.8)

0.006

1.2 (0.7 - 1.9)

0.510

treatment only

 

Smoking (%)g

45.8

0.2

NA

1.0 (0.7 - 1.3)

0.870

0.9 (0.7 - 1.2)

0.370

none

 

Syncopal episodes (%)g

16.6

23.5

17.7

0.5 (0.3 - 1.0)

0.036

2.4 (1.5 - 3.9)

<0.001

both

x

  1. NA Not applicable as variables were recorded as non-time-varying
  2. aPercentage of individuals reporting at least one missing data point per time-varying characteristic or completely missing in non-time-varying characteristics given a complete cohort N of 639 individuals
  3. bPercentage of individuals reporting time-varying data per characteristic (e.g. “Marital status” changed at least once over the course of time in 6.6% of the cohort)
  4. cOdds ratio for the association between “physical activity” and characteristic over all time-points (taking individual-specific clustering into account); higher ratios indicate more likely to be “physically active” given the characteristic or per unit increase in characteristic
  5. dHazard ratio for the association between “physical activity” and characteristic; higher ratios indicate more likely to transit to disability given the characteristic or per unit increase in characteristic
  6. eAssociations are highlighted; note that an association is arbitrarily defined as being “relevant” if the univariate regression model is significant to p < 0.1 (see the Supporting information for details of why this cut-off is chosen)
  7. fLevel of education: 10 years of education or less of formalized education (reference) vs. more education; Degree of ARWMC rated according to “Fazekas criteria”; Stroke: recorded stroke event independent of any severity post-baseline; Alcohol consumption: any alcohol consumption, including for the largest part casual drinking (reference) vs. none; Anxiety or depressed mood: self-reported; Body mass index: baseline tertiles - change is in respect to baseline. Major depressive episode: DSM-IV expert rating; MMSE: >24 points (reference) vs. worse performance
  8. gCharacteristics are not present (reference) vs. they are