Introduction In 2014, the contract to deliver the National Lung Cancer Audit (NLCA) was awarded to the Royal College of Physicians. Data were previously submitted using a bespoke dataset (LUCADA), but will now be submitted via the nationally mandated Cancer Outcome and Services Dataset (COSD) and linked to additional cancer registry datasets. For patients diagnosed in 2014, NLCA data were submitted using LUCADA for 132 of 151 English trusts. Trusts also submitted data via COSD and registry data were produced by the National Cancer Registration Analysis Service (NCRAS), providing the opportunity to compare both datasets for data completeness and reliability.

Methods We have linked the LUCADA and cancer registration datasets at patient level and assessed completeness of key patient variables including age, sex, stage, performance status and pathological confirmation, as well as recording/dates of treatment received. We assessed the inter-rater/data agreement of these variables using Cohen’s kappa statistics (k). Finally, we carried out a qualitative assessment on a subset of cases to explore reasons why patients were represented in one dataset but not the other.

Results There were 26,001 patients in both datasets (94% of LUCADA data) with more in the registry dataset and not LUCADA than vice versa. Recorded sex and age were highly congruent, as was trust first seen which was the same in 96%. 56% of the patients had the same date of diagnosis, 74% were ± 7 days and 86% were ± 14 days of each other. The cancer registry data had a larger proportion of patients with missing PS (27% vs 11%) with agreement on PS (where available) being 97% (k = 0.91). Agreement on stage was 94% (k = 0.81). Agreement for surgery, chemotherapy and radiotherapy was 0.86, 0.88 and 0.77 respectively. Details of the qualitative work and trust first seen algorithms will be provided in the presentation.

Conclusion Results suggest that cancer registry data accurately describe key patient features. Compared with LUCADA, the national cancer registry:

• identified more patients

• has a higher proportion pathological confirmation

• identified more patients with surgery, chemotherapy and radiotherapy

• has a higher proportion of missing data for PS which could be due to data entry transition