Agreement Between Self-Reported Information and Health Register Data on Chronic Diseases in the Oldest Old.

Abstract

To study the agreement on disease prevalence between survey data and national health register data among people aged over 90. The survey data were from the Vitality 90+ Study conducted among 1637 community dwellers and persons in long-term care aged 90 and over in Tampere, Finland. The survey was linked with two national health registers, including hospital discharge data and prescription information. The prevalence of 10 age-related chronic diseases was calculated for each data source and the agreement between the survey and the registers was estimated using Cohen's kappa statistics and positive and negative percent agreement. The prevalence of most diseases was higher in the survey than in the registers. The level of agreement was highest when the survey was compared with information combined from both registers. Agreement was almost perfect for Parkinson's disease (ĸ=0.81) and substantial for diabetes (ĸ=0.75) and dementia (ĸ=0.66). For heart disease, hypertension, stroke, cancer, osteoarthritis, depression, and hip fracture, the agreement ranged from fair to moderate. Self-reported information on chronic diseases shows acceptable agreement with health register data to warrant the use of survey methods in population-based health studies among the oldest old. It is important to acknowledge the gaps in health registers when validating self-reported information against register data.