Table 3

Adjusted* Probit regression estimates work status change due to chronic disease diagnosis at time 1 and time 2

EstimateSEp Value
Arthritis (N=3874)
 Time 1 total effects (path A)0.2000.055<0.001
 Time 2 total effects (path C+(path A×path B))0.3600.090<0.001
 Time 2 total effect estimate
  Direct effect (path C)0.1070.0920.246
  Effect mediated through time 1 work status (path A×path B)0.2540.0760.001
 Model fit: CFI=0.964, TLI=0.927; RMSEA 0.013 (0.006–0.018)
Back problems (N=4655)
 Time 1 total effects (path A)0.1940.0610.001
 Time 2 total effects (path C+(path A×path B))0.2120.0900.019
 Time 2 total effect estimate
  Direct effect (path C)−0.0320.1030.753
  Effect mediated through time 1 work status (path A×path B)0.2440.0820.003
 Model fit: CFI=0.997, TLI=0.994; RMSEA 0.003 (0.000–0.010)
Diabetes (N=1085)
 Time 1 total effects (path A)0.3210.081<0.001
 Time 2 total effects (path C+(path A×path B))0.3600.1090.001
 Time 2 total effect estimate
  Direct effect (path C)−0.0260.0910.772
  Effect mediated through time 1 work status (path A×path B)0.3870.104<0.001
 Model fit: CFI=0.996, TLI=0.990; RMSEA 0.010 (0.000–0.028)
Hypertension (N=3120)
 Time 1 total effects (path A)0.0640.0680.341
 Time 2 total effects (path C+(path A×path B))0.1390.0880.112
 Time 2 total effect estimate
  Direct effect (path C)0.0520.1010.605
  Effect mediated through time 1 work status (path A×path B)0.0870.0930.349
 Model fit: CFI=0.991, TLI=0.981; RMSEA 0.006 (0.000–0.015)
Heart disease (N=883)
 Time 1 total effects (path A)0.7050.107<0.001
 Time 2 total effects (path C+(path A×path B))0.3870.1560.013
 Time 2 total effect estimate
  Direct effect (path C)−0.5970.3630.100
  Effect mediated through time 1 work status (path A×path B)0.9830.3440.004
 Model fit: CFI=0.993, TLI=0.985; RMSEA 0.008 (0.000–0.027).
  • Path model corresponds to figure 1, and is stratified by a chronic health condition of interest. Positive estimates indicate a higher likelihood of not working due to health reasons. A positive and larger estimate indicates an increase in the likelihood of an outcome.

  • *Estimate adjusted for baseline measures of sex, age (grouped) comorbidities, body mass index (grouped), occupational strength requirements, self-reported depression, marital status, geographic location, survey year and work hours. Based on the process of selecting respondents, each model may consist of a different subsample of the overall cohort and may to some extent overlap.

  • CFI, comparative fit index; RMSEA, root mean square error of approximation; TLI, Tucker-Lewis index.