The purpose of this project was to compare three hospital-based measures of co-morbidity to patient self-report co-morbidity and to determine the relative proportion of outcome predicted by each of the co-morbidity measures in a population of individuals receiving major joint arthroplasty. Baseline measures using the SF-36 general health questionnaire and the Western Ontario McMaster Osteoarthritis Index (WOMAC) were obtained from 518 persons undergoing total knee or hip replacement. A second measure was obtained six months post-surgery. Co-morbidity was calculated by summing the self-reported co-morbidity at baseline, using both the chart-based and administrative data version of Charlson's Co-morbidity Index, and by summing the number of International Classification of Diseases-Version 9 (ICD-9) codes appearing in the electronic health record. Linear regression was used to determine how much of the variation in outcome was explained by each of the co-morbidity measurement methods. Self-report co-morbidity explained as much variation in outcome as the hospital-based indices. Self-report co-morbidity did not perform as well as the other methods in explaining the variance in health system utilization. It was concluded that self-report co-morbidity is minimally as useful as record-based systems when measuring the impact of co-morbidity on health related quality of life (HRQL). This is an important finding, as privacy legislation, the time until data is available and cost are all barriers to using record-based co-morbidity measures.