Predictors of diagnostic interval and associations with outcome in acute lymphoblastic leukemia

BACKGROUND: Little is known about diagnostic interval lengths in childhood cancer, their predictors or impact upon survival. To date, studies have relied on questionnaires or chart abstraction. We aimed to construct and validate a diagnostic interval measure using health services data among children with acute lymphoblastic leukemia (ALL) in order to determine predictors of prolonged intervals and associations with event-free survival (EFS). PROCEDURE: All children with ALL diagnosed 1995-2011 (N = 1,541) in Ontario, Canada were linked to population-based health administrative databases. Healthcare claims prior to diagnosis were used to define healthcare episodes. Diagnostic intervals (time between first episode with diagnostic code a priori classified as consistent with underlying ALL, and diagnosis) were validated by correlation with a chart abstraction-based measure. RESULTS: Intervals were generally short (median 2 days, IQR 1-3). Predictors of longer intervals included having general primary care physicians versus pediatricians (odds ratio 1.60, 95%CI 1.04-2.47; P = 0.03). While prolonged diagnostic intervals were associated with superior EFS (hazard ratio 0.71, 95%CI 0.52-0.98; P = 0.04), this was explained by confounding by disease biology. CONCLUSIONS: Health administrative data can be used to measure diagnostic intervals in ALL and potentially other pediatric malignant and non-malignant diseases. Diagnostic intervals were short and a marker of disease severity rather than independent predictors of outcome. These findings may be used to address caregiver guilt and caution against "early diagnosis" benchmarks not based in evidence. Future studies should examine the impact of diagnostic interval length in other conditions, but should account for potential confounding by disease severity.