³Ô¹ÏÌìÌÃ91

Workers/General Population: (Turpin et al, 2004; n = 7,797)

• Significant main effect for sex (p < 0.05) and WIS: Men (20.6) reporting marginally less impaired presenteeism than women (21.3).
• Significant main effect for age and WIS (p < 0.05) showcasing a trend for less impaired presenteeism being reported as employee¡¯s age increases (WIS scores ranged from 22.2 at less than 25 years of age, to 19.8 at age 56 and older.)
• Significant difference between the WIS reported for persons with arthritis/joint pain in knowledge-based jobs (18.67) and persons with arthritis/joint pain in production-based jobs (22.48; t = 3.175; p < 0.002). The magnitude of the difference was fairly large (about 20%).
  • Large, but statistically insignificant differences were observed for other primary health conditions such as asthma, most likely due to small cell sizes.

Employees at Two Enterprises: (Yamashita & Arakida, 2008, First survey to 359 female employees in enterprise A with inclusion criteria of presence of some health conditions and answers to all 10 items on the WIS; age groups from 20-29 to 50+; 187 (52.1%) full-time; job type = clerks (51.0%), production-based workers (39.4%), technicians (3.1%), and others (6.5%). Second test-retest survey to 50 female employees (33 responses) with inclusion criteria the presence of some health conditions at both time points, mean age = 40.6 (9.8), 80% were occupational health staff. Japanese sample.)

Mean scores of WIS, WOS, and hours of sickness absenteeism (n = 359)

Work impairment scores by age, all subjects

F = 5.3, p < 0.01

Work impairment scores by age, full-time workers

F = 0.85, p > 0.05

Work impairment scores by age, part-time workers*

F = 2.71, p > 0.05

*ANOVA among part-time workers performed by excluding a subject less than 30 years old.

Employees at Two Enterprises: (Yamashita & Arakida, 2008, n = 33, Japanese sample)

• Acceptable test-retest reliability for the WIS: (ICC = 0.86; r = 0.83, p < 0.001)
• Acceptable test-retest reliability for the Work Output Score (WOS): (ICC = 0.91; r = 0.789, p < 0.001)  

Workers/General population: (Turpin et al., 2004); n = 7,797)

• Excellent: Cronbach's alpha coefficient for the WIS was 0.82*.

*Reliability improved marginally to 0.83 in overall population and types of job when an item was removed (¡°were you unable to take satisfaction in your work¡±).

Employees at Two Enterprises: (Yamashita & Arakida, 2008, n = 359, Japanese sample)

• Excellent: Cronbach's alpha of WIS was 0.87
• Excellent: Cronbach's alpha of WIS in full-time workers was 0.88
• Excellent: Cronbach's alpha of WIS in part-time workers was 0.85

Predictive validity:

Workers/General Population: (Turpin et al, 2004; n = 7,797)

• Poor predictive validity between health care use/expenditures and absenteeism or presenteeism metrics (r < 0.10 for all) 
• Significant and robust positive relationship between the WIS and number of hours of absenteeism self-reported on the SPS 
• Correlations between administrative-report payroll absenteeism and WIS were not significant 

Concurrent validity:

Workers/General Population: (Turpin et al, 2004; n = 7,797)

• Adequate concurrent validity between WIS and the Work Limitations Questionnaire (WLQ) Overall Index (r = 0.50) 
• Poor to adequate concurrent validity between WIS and the four subscales of the WLQ: output demands scale (r = 0.49), mental/ interpersonal demands (r = 0.41), physical demands (r = 0.37), and time management (r = 0.20). 

Construct Validity:

Workers/General Population: (Turpin et al., 2004); n = 7,797)

• SPS appeared to be valid for both knowledge- and production-based jobs, yet there are important differences between them. The first factor of WIS for knowledge-based jobs (Completing Work) accounted for 93% of the variance in WIS scores, while the second factor of WIS (Avoiding Distraction) accounted 95% of variance in WIS scores in production-based jobs. Thus, all ten items of WIS are essential to assess and account for impairments in presenteeism.

Employees at Two Enterprises: (Yamashita & Arakida, 2008, n = 359, Japanese sample)

• Factorial analysis of the WIS data identified two factors: ¡°completing work¡± and ¡°avoiding distraction.¡±
  • These two factors accounted for a cumulative 50.8% of the variance¡ª49.9% for full-time workers and 51.9% for part-time workers

Convergent Validity:

Workers/General population: (Turpin et al., 2004); n = 7,797)

• Convergent validity was assessed through the comparison of the WIS and WOS with health status
• WIS was significantly correlated to all eight SF-36 subscales: physical functioning (r = -0.25), role limitations-physical (r = -0.35), pain (r = -0.33), general health (r = -0.45), vitality (r = -0.61), social functioning (r = -0.56), role limitations-emotional (r = -0.51), and general mental health (r = -0.62) 
  • Some differences between production- and knowledge-based jobs on the above correlations, but the trends are similar. 
• WLQ had lower correlations with the SF-36 subscores in comparison to the SPS (total population range: 0.31-0.44; knowledge-based jobs range = 0.15 to 0.40, higher correlations on mental health questions; production-based jobs range = 0.29 to 0.56

Employees at Two Enterprises: (Yamashita & Arakida, 2008, n = 359, Japanese sample)

• Poor to Adequate convergent validity between the Work Impairment Score (WIS) and The Medical Outcomes Study 36-item Short Form Health Survey (SF-36)

Pearson¡¯s correlation coefficients between WIS and SF-36 subscores

^ap < 0.05 (Poor), ^bp < 0.01 (Adequate)

• Excellent convergent validity between WIS and WOS (r = -0.66**) 
• Poor convergent validity between WIS and Job Content Questionnaire subscores for demand (r = 0.13*), support (r = 0.22**), and daily hassles (r = 0.29**) 
• Adequate negative correlations between WIS and job satisfaction (r = -0.39**) 

*p < 0.05, **p < 0.01

Musculoskeletal Conditions: (Roy et al., 2011)

• Excellent: Cronbach¡¯s alpha ranged between 0.82 and 0.87 for the Work Impaired Score (WIS) subscale.

Convergent validity:

Musculoskeletal Conditions: (Roy et al., 2011)

• Low to moderate correlations between SPS-13, WLQ-25, and SF-36 were observed. 
• Adequate to excellent correlations were obtained between the presenteeism scales of EWPS, SPS-13, AND WLQ-25 (r = 0.55 to 0.69)