Abstract:
Studies of work-related musculoskeletal diseases frequently rely on questionnaires for the assessment of physical exposures, yet the reliability and validity of these measures are often unknown. We studied the reliability and validity of a questionnaire which assessed occupational physical exposures to the back and lower extremities.
A convenience sample of retired workers (mean age 69 years) was chosen for a test-retest study of questionnaire reliability. 36 out of 53 subjects completed a Job Physical Demands questionnaire on two occasions six weeks apart. For all past jobs held, this 9-item scale estimated physical exposures such as walking, lifting, and climbing, as well as overall physical work demands.
For individual past jobs, test-retest of job strength ratings showed excellant reliability (intraclass correlation coefficient [ICC]=0.84). Test-retest correlations for individual exposure variables ranged from poor (ICC<0.40) to excellant (ICC>0.75), with better scores obtained when subjects used a categorical rather than continuous format to rate exposures. Estimates of lifetime physical exposures showed good to excellant reliability, withh ICC of >0.75 seen for lifetime job strength ratings, years of heavy or very heavy work, lifetime pounds lifted, and lifetime climbing stairs or ladders.
There was good correlation (ICC=0.67) between self-reported exertion and the strength requirement listed by the US Department of Labor Dictionary of Occupation Titles, supporting the validity of recall on the questionnaire. The validity fo self-reported exposures was further tested in a group of 120 actively employed workers at an aluminum reduction plant who completed a Job Physical Demands questionnaire and had ergonomic analysis of their current jobs. Self-reports showed good correlation with observed values (r2=0.80), and discriminated appropriately between two different jobs in 9 of 11 exposure categories.
These data show that questionnaires may provide valid measures of some physical exposures, and are reliable even for exposures which occured in the distant past. Large differences in reliability were found using different questionnaire formats, with results favoring ordinal rather than continuous response formats.