Interviews or medical records, which type of data yields the best information on elderly people's health status?

Background and aims: Self-reported data and/or medical records are often used to assess the prevalence of illness and impairment in epidemiological studies. However, these two data sources do not always provide the some information. The aim was to compare data from interviews and medical records regarding illness, symptoms and impairment in the elderly, and to analyze the agreement between a consensus from both data sources and data from interviews and medical records, respectively. Methods: We interviewed 130 persons (age range 67-99) regarding socio-demographic background data and physical and mental health. Medical records were reviewed. Illness burden was rated according to the Cumulative Illness Rating Scale for Geriatrics, and was rated in three ways based on: 1) interview data; 2) medical records; 3) information from both interviews and medical records considered to be consensus. Agreement was measured by the Kappa coefficient and the Svensson Paired Rank Measurement. A permutation test tested whether the ratings from interviews and medical records had the same agreement when compared with consensus. Results: Statistically significant differences in agreement were found between interview versus consensus and medical records versus consensus for the vascular system (medical records best), eyes/ears/ nose/throat/larynx and musculoskeletal/integument (interview best). Medical records gave better information concerning specific diseases and diagnoses, whereas interview data provided a better measure of illness, functional impairment and health in a broader sense. Conclusions: Both medical records and interviews yield good information of elderly people's health status, but they focus on different aspects of health.