Comorbidity is an important dimension of patient health status. However, limited attention has been given to assessing the reliability of patient-reported data on comorbid conditions.
We sought to evaluate the consistency of self-report of 12 comorbid conditions at 3 distinct time points and examine the sociodemographic, clinical, health status, and survey factors associated with reliability.
We undertook a longitudinal cohort analysis of survey and medical record abstract data obtained from a population-based sample of 3095 prostate cancer survivors.
Consistent and inconsistent response patterns were assessed using descriptive statistics and regression modeling.
More than half of the cohort demonstrated consistent responses to all comorbid condition questionnaire items. Arthritis had the highest (13%) and liver disease the lowest (1%) proportion of inconsistent responses. Older age, lower income, and poorer general and mental health status were significant predictors of inconsistent responses. Subset analyses of the 4 most prevalent comorbid conditions (ie, arthritis, diabetes, hypertension, and depression) showed that respondents who reported that they were limited by, or taking prescription medicine for, the condition were more likely to provide consistent responses than those neither limited nor taking medicine. Response consistencies of 92% or better were obtained for 11 of the 12 conditions.
Men with a relatively recent prostate cancer diagnosis are generally able to provide reliable reports of their concomitant health conditions. To increase the likelihood of obtaining reliable data, investigators should consider ascertaining condition severity and current medical management when querying subjects about comorbid conditions in surveys.