Validity of using general and cancer-specific health administrative data to identify treatment exposures in adolescents and young adults with cancer: A population-based study using the IMPACT cohort.
AI interpretation is pending for this paper.
Open original publication →What the AI sees
Not AI summarized yet.
Research significance
Pending deeper interpretation.
Source abstract
INTRODUCTION: Cancer treatment details (i.e., radiation site, chemotherapy dose) are required to conduct rigorous health services research but are difficult to obtain from administrative data. We conducted a validation study to ascertain the optimal algorithm for defining cancer treatment details in cancer-specific and general health administrative data using available chart-abstracted data from adolescent and young adult (AYA) cancer patients. METHODS: Health administrative data and cancer treatment data reported by visit in Ontario, Canada were compared separately and in a combined algorithm to a reference-standard chart-abstracted database of AYA cancer patients diagnosed in 2005-2012 (n = 1173). We tested algorithms for three tiers of treatment details: any chemotherapy/radiotherapy provided; type of chemotherapy/site of radiation; dose of chemotherapy/radiation. For each algorithm, we calculated sensitivity, specificity, positive predictive value, negative predictive value with 95 % confidence intervals (95 %CI) and simple kappa statistics, overall and according to cancer type, diagnosis period, and locus of care. RESULTS: General health administrative data had high sensitivity and specificity (> 80 %) for detection of any chemotherapy (n = 942) or radiation exposure (n = 412) and was not improved by using cancer-specific data. In 475 patients (40.5 %) with chemotherapy treatment records, sensitivity (22.4-59.6 %) and specificity (95.8-99.1 %) varied by chemotherapy type/class. Factors associated with missing records include locus of care (9.5 % in pediatric vs. 81.7 % in adult cancer centres), year of diagnosis, and type of cancer. There was moderate to strong correlation (r = 0.50-0.79) between dosing for the most common anthracyclines, combined alkylators, cisplatin, and bleomycin. For radiation treatment data (n = 406, 98.5 %), sensitivity and specificity for radiation site ranged from 73.4 % to 91.2 % and 96.6 % to 99.7 %, respectively, with strong dosing correlation (r = 0.63-0.95, by site). CONCLUSIONS: Both general and cancer-specific health administrative data have value in determining receipt of chemotherapy and/or radiation and can be used reliably to create cohorts of exposed cancer patients. More granular information regarding dose and type of chemotherapy and dose and site of radiation therapy is highly specific but limited by variable sensitivity. Care should be taken when using the data to estimate prevalence, compare treated/untreated groups or when full capture of an exposed population is otherwise required as underestimations of the true effect may occur.