Agreement between individual and neighbourhood income measures in patients with colorectal cancer in Canada
Menée au Canada à partir de données portant sur 103 530 patients atteints d'un cancer colorectal diagnostiqué entre 1992 et 2017, cette étude identifie les erreurs de classification du revenu (revenu individuel et revenu du voisinage) puis analyse l'association entre ces erreurs et la survie globale
Introduction: With increasing interest in income-related differences in cancer outcomes, accurate measurement of income is imperative. Misclassification of income can result in wrong conclusions as to the presence of income inequalities. We determined misclassification between individual- and neighbourhood-level income and their association with overall survival among colorectal cancer (CRC) patients. Methods: The Canadian Census Health and Environment Cohorts were used to identify CRC patients diagnosed from 1992 to 2017. We used neighbourhood income quintiles from Statistics Canada and created individual income quintiles from the same data sources to be as similar as possible. Agreement between individual and neighbourhood income quintiles was measured using cross-tabulations and weighted kappa statistics. Cox proportional hazards and Lin’s semi-parametric hazards models were used to determine the effects of individual and neighbourhood income independently and jointly on survival. Analyses were also stratified by rural residence. Results: 103,530 CRC patients were included in the cohort. There was poor agreement between individual and neighbourhood income with only 17% of respondents assigned to the same quintile (weighted kappa=0.18). Individual income had a greater effect on relative and additive survival than neighbourhood income when modeled separately. The interaction between individual and neighbourhood income demonstrated that the most at risk for poor survival were those in the lowest individual and neighbourhood income quintiles. Misclassification was more likely to occur for patients residing in rural areas. Conclusion: Cancer researchers should avoid using neighbourhood income as a proxy for individual income, especially among patients with cancers with demonstrated inequalities by income.