Response stability and reliability in longitudinal health evaluations

Two approaches were used to study the stability over time and intravisit reliability of health questions and clinical medical examination items in the Baltimore Longitudinal Study of Aging (BLSA). The stability of responses was determined by evaluating the medical history and physical examination completed at each visit to identify items that once answered in a positive manner, should continue to be answered positively over time. Stability for each question and subject was calculated by the number of positive responses following the first positive response divided by the total number of visits following the first positive response. For 35 questions answered by the subject, the stability was 58% by a simple average or 64% when weighted for the percentage of subjects who had a positive response to the question; for 10 physician-asked questions, the corresponding figures were 29% or 36%. Eighteen items from the physical examination had a stability of 34% or 37%. Intravisit reliability was estimated by comparing responses from the general health questionnaire to responses on the Cornell Medical Index completed at the same visit. Subject-completed questions had substantial agreement (Kappa=0.74, for questions worded the same), while physician-asked questions had moderate agreement (Kappa=0.44). (Aging Clin. Exp. Res. 4: 43–52, 1992)