Statistical 'biases' in respiratory disability determinations

The manner in which pulmonary function test results are employed in the assessment of respiratory disability may be affected by 4 statistical choices: (1) choice of prediction equation(s), (2) adjustment factors (such as sex and race), (3) criterion values, and (4) method of comparison of observed to predicted normal test values. The records of 900 respiratory disability applicants were employed to estimate the direction and magnitude of the effect of these choices on the overall number of persons who would be declared 'disabled' and upon the manner in which personal characteristics (e.g., sex, race, height, age) affect the likelihood of being declared 'disabled'. Choice of prediction equation had minor effects, and adjustment for race and sex had more significant effects. Choice of criterion value affected the overall number and, in certain instances (e.g., Social Security Disability Insurance), the distribution of 'disability' declarations. Method of comparison (percent of predicted, difference of predicted minus observed or minimal value criterion) had major effects upon the distribution of 'disability' declarations between population subgroups. Preliminary analysis therefore suggests that these statistical choices should be carefully manipulated in the design of a disability system to facilitate achievement of the system's goals.