Physical Activity components (algorithm)

Physical Activity components (algorithm)

LASA filenames:
LASA046

Contact: Bram Hoogerheide


Background
Physical activity (PA) in research is often measured using minutes per day or intensity measures like metabolic equivalent task (MET) scores, which quantify energy expenditure relative to resting levels. In LASA, these measures are assessed using the LASA Physical Activity Questionnaire (LAPAQ)[1] and further described in the Physical activity documentation.

However, PA is a very broad concept with complex relationships to health that cannot fully be captured in only duration and/or intensity. For example, activities with high mechanical strain increases the risk of osteoarthritis, while activities requiring high lower body muscle strength decreases this risk of osteoarthritis[2].


PA components and algorithm
Based on literature review, expert consensus, and available LASA data, an algorithm for the calculation of PA component scores of 1) intensity, 2) mechanical strain, 3) turning actions, and 4) muscle strength has been developed[3]. PA components scores are based on the following LAPAQ activities: walking outdoors, cycling, light household activities, heavy household activities, and two of the most frequently performed sports. For each activity, component scores were assigned.

Intensity scores consist of adjusted MET scores for older adults, with one MET corresponding to 1kcal*kg body weight*hour. This resulted in a component score ranging from 2,5 (e.g. for the activity outdoor walking) to 6,0 (e.g. for the sports activity tennis).

Mechanical strain scores are based on ground reaction forces, and reflect the loading on the bones in multitudes of bodyweight. A score of 1 was assigned to non-bearing activities and standing activities, a score of 2 to weight bearing activities, a score of 3 to activities that include explosive actions such as sprinting and running and a score of 4 to activities that include jumping. The mechanical strain component therefore ranges from 1 to 4.

Turning actions reflect the amount of transversal rotations of the lower extremities, with scores ranging from 1 to 3, with the highest scores reflecting activities with many rotations, such as dancing and tennis.

Strength scores indicate the amount of lower body muscle strength to perform an activity, and range from 1 to 4. A score of 1 was assigned to standing and sitting activities, a score of 2 was assigned to activities related to walking, a score of 3 was assigned to activities involving running and cycling, and a score of 4 was assigned to activities that include jumping and squatting.

Duration was later added as a fifth component[5] for easier comparisons and interpretations of the measurements over time, and ranges from 0 to 960 minutes per day (0 to 16 hours). This component reflects the mean time spent per day on performing physical activities.

Scores for each activity performed were subsequently averaged into one score per component per wave, and finally new variables are created to cut each averaged PA component score into tertiles of ‘low’, ‘medium’, and ‘high’ scores to increase the contrast. When using tertiles of PA component scores in longitudinal research, it is important that the cut-off values remain exactly the same over all waves, so individuals can change relative to their baseline values. Within the algorithm, the cut-off values from the B wave are applied to all the waves.

The algorithm is based on data from the B wave, but can be applied to all subsequent waves. Currently, SPSS syntax is only available for wave B, but can be easily adjusted for other waves by changing the corresponding variable names and by applying the activity recoding for waves 2B and newer (see also the  documentation file in PDF). R code is directly available for data in a long format (so all waves merged into one longitudinal dataset), but can also be applied to specific waves after changing the variable names.

So variable “lphya01” is named “blphya01” for wave B, “clphya01” for wave C, etc. Variables in the SPSS syntax are all written in the form “blphya01” to correspond with wave B. Variables in the R code are all written in “lphya01”, and thus will need some adjustments for applying this code cross sectionally. More detailed information on the merging of waves in a long format in R, and construction and evaluation of the PA components can be found in Hoogerheide B (Bram). LASA PA component documentation 2025. https://doi.org/10.17605/OSF.IO/9U3EY. More information on the components and their development is reported by Verweij et al[3]. These components have been found to associate differently with the risk of falling[4] and incidence of osteoarthritis[2], and can thus be a meaningful addition to common measures of duration and intensity.


Data sources
Data files and variables used in PA component algorithm:

  • Physical Activity: LASA046


Availability of information per wave¹

B C D E 2B F G H 3B I J K
Total N 3107 3107 3107 3107 1002 4109 4109 4109 1023 5132 5132 5132
Complete 2516 1937 1388 1086 683 1496 1209 1207 978 1685 1354 1089
Incomplete 354 246 336 249 291 388 367 84 39 67 32 13
Invalid 45 39 18 22 21 23 19 15 4 16 6 8
Missing 192 469 591 634 7 626 626 648 2 826 936 948
Deceased 0 416 774 1116 0 1576 1888 2155 0 2538 2804 3074

¹ These numbers are cross-sectional and thus not longitudinally cleaned.


Complete records reflect the number of respondents per wave with all 14 LAPAQ items present used in constructing the PA component scores. Incomplete records have at least one missing response on these 14 items. Invalid records are the number of respondents per wave who are bedridden or bound to wheelchair at the current wave. Missing records are completely missing LAPAQ records, due to shortened or terminated interviews, or due to non-participation in the current wave or study drop-out. Note that additional cross-sectional cleaning might be required to remove invalid measures (e.g. total PA duration more than 24 hours per day).

Incomplete records can be imputed, as well as the missing records due to the shortened or terminated interview. It is recommended that at least Z004 (sex) and Z008 (age at interview) are used for imputing missing records (although using more variables will result in better imputations), and Z990 (mortality data) and Z002 ((non-)response data) to select the right records for imputing. Missing records due to study drop-out or non-participation should be considered missing not at random, and should not be imputed.


Previous use in LASA

  • Verweij LM, Van Schoor NM, Deeg DJH, Dekker J, Visser M. Physical activity and incident clinical knee osteoarthritis in older adults. Arthritis Care Res (Hoboken) 2009;61:152–7. https://doi.org/10.1002/art.24233.
  • Verweij LM, van Schoor NM, Dekker J, Visser M. Distinguishing four components underlying physical activity: a new approach to using physical activity questionnaire data in old age. BMC Geriatr 2010;10:20. https://doi.org/10.1186/1471-2318-10-20.
  • Peeters GMEE, Verweij LM, Van Schoor NM, Pijnappels M, Pluijm SMF, Visser M, et al. Which types of activities are associated with risk of recurrent falling in older persons? Journals of Gerontology – Series A Biological Sciences and Medical Sciences 2010;65 A:743–50. https://doi.org/10.1093/gerona/glq013.
  • Hoogerheide B, Maas ET, Visser M, Hoekstra T, Schaap LA. Trajectories of physical activity components among community-dwelling older adults. In Press n.d.


References

  1. Stel VS, Smit JH, Pluijm SMF, Visser M, Deeg DJH, Lips P. Comparison of the LASA Physical Activity Questionnaire with a 7-day diary and pedometer. J Clin Epidemiol 2004;57:252–8. https://doi.org/10.1016/j.jclinepi.2003.07.008.
  2. Verweij LM, Van Schoor NM, Deeg DJH, Dekker J, Visser M. Physical activity and incident clinical knee osteoarthritis in older adults. Arthritis Care Res (Hoboken) 2009;61:152–7. https://doi.org/10.1002/art.24233.
  3. Verweij LM, van Schoor NM, Dekker J, Visser M. Distinguishing four components underlying physical activity: a new approach to using physical activity questionnaire data in old age. BMC Geriatr 2010;10:20. https://doi.org/10.1186/1471-2318-10-20.
  4. Peeters GMEE, Verweij LM, Van Schoor NM, Pijnappels M, Pluijm SMF, Visser M, et al. Which types of activities are associated with risk of recurrent falling in older persons? Journals of Gerontology – Series A Biological Sciences and Medical Sciences 2010;65 A:743–50. https://doi.org/10.1093/gerona/glq013.
  5. Hoogerheide B, Maas ET, Visser M, Hoekstra T, Schaap LA. Trajectories of physical activity components among community-dwelling older adults. In Press n.d.

 

Date of last update (first version): September 9, 2025