Objective: To determine the relationship between regional variations in knee replacement (KR) utilization rates in Ontario, Canada, and the reported prevalence of arthritis and rheumatism as a chronic health problem.
Methods: Utilization data were acquired from the Canadian Institute for Health Information for KR procedures performed in Ontario between fiscal years 1984 and 1990. Census information was obtained from Statistics Canada. Disease prevalence data were derived from the 1990 Ontario Health Survey (OHS). Public Health Units (PHU) were used as the unit of analysis, with utilization rates defined as the number of KR performed on all PHU residents (irrespective of where these procedures were performed) divided by the population. Direct methods were used to standardize utilization for age, sex, and disease prevalence. The extremal quotient, the weighted coefficient of variation, and the systematic component of variation were used as measures of variation. The relationship between the number of KR performed in each age-sex-year strata and various demographic (age and sex), disease prevalence, and regional dummy variables was estimated using a Poisson regression model.
Results: Regional variation in the standardized utilization of KR surgery was wide, but declined over the study period; the extremal quotient fell from 8.0 to 3.3, the weighted coefficient of variation fell from 0.49 to 0.30, and the systematic component of variation fell from 0.20 to 0.17. Variation in the provision of KR surgery remained even after controlling for the demographic composition of the population and disease prevalence. Moreover, while demographic, regional, and temporal covariates were significant (p < 0.0001) in accounting for over 90% of the variation in utilization, disease prevalence was not significant (p > 0.05).
Conclusion: This study merged population based reports of disease prevalence with administrative data to account for regional variations in utilization. While regional variations in KR surgery have fallen over time, variations remain even after adjusting for patient reported disease prevalence. The finding that demographic variables and the reported prevalence of disease were poorly correlated suggests that current area variation studies may not be adjusting fully for disease prevalence or severity.