Table 2

Stability of physical activity variables over time, n=3413

Waves 1–2Waves 1–3Waves 1–4Random effects models
UnivariateMultivariate*
KappaKappaKappaICC (95% CI)ICC (95% CI)
Physically active0.260.230.240.46 (0.43 to 0.48)0.44 (0.41 to 0.46)
Sport participation0.380.350.350.65 (0.63 to 0.67)0.61 (0.59 to 0.63)
Recreational activity0.240.190.160.38 (0.36 to 0.40)0.36 (0.34 to 0.39)
Walking0.150.110.120.32 (0.30 to 0.35)0.32 (0.30 to 0.34)
  • Kappa statistics are presented for participants with a valid physical activity score at all four time points (n=3413). Data on walking were missing for an additional 15 participants at 12-year follow-up, 3 participants at 16-year follow-up and 1 participant at 20-year follow-up. Data on sport participation were missing for 33 participants at 12-year follow-up, 68 participants at 16-year follow-up and 34 participants at 20-year follow-up. Random effects models included men with at least two assessments for each domain accompanied by valid covariate data (physical activity: n=5962; sport participation: n=6122; recreational activity: n=6093; walking: n=6040). ICC from random effects models.

  • *Adjusted for age, body mass index, social class and smoking status at baseline.

  • ICC, intraclass correlation coefficients.