Background Primary care datasets offer valuable longitudinal data for research and policy-making. However, Ireland’s primary care data infrastructure is limited, with inconsistent diagnostic coding raising concerns about research utility. While gaps in coding chronic conditions are well-documented, cancer diagnosis validation remains unaddressed. This study assesses the utility of Irish GP data by comparing cancer incidence rates from primary care records with the National Cancer Registry Ireland (NCRI). Methodology A retrospective cohort study used anonymised data from 43 GP practices in Ireland, covering the period from 1 January 2011 to 5 April 2018. Cancer cases were identified using ICPC-2 and ICD-10 diagnostic codes for the 20 most common reported cancers by NCRI. Age- and sex-adjusted cancer incidence was derived from NCRI data and compared with observed GP-recorded cases using standardised incidence ratios (SIRs). Chi-squared tests compared observed and expected frequencies. Inter-practice variation was assessed by comparing expected and observed case counts. Results The cohort comprised 41,782 patients aged ≥60 years, with mean follow-up of 5.3 years. Across 20 cancers examined, 15 were significantly under-recorded in GP data relative to NCRI estimates. No significant differences were observed for colorectal, leukaemia, thyroid, cervical cancers, or Hodgkin’s lymphoma. In contrast, marked under-recording was observed for melanoma, liver, pancreatic, brain, and ovarian cancers. Prostate (SIR 1.20), breast (SIR 1.53) and bladder cancer (SIR 2.14) were significantly over-recorded. Cancers lacking granular ICPC-2 diagnostic codes and relying on ICD-10 demonstrated the greatest under-ascertainment. Inter-practice variation was substantial with marked heterogeneity in coding and wide dispersion in practice-level SIRs. Conclusion Irish GP datasets substantially under-record cancer incidence versus NCRI estimates, primarily due to heterogeneity in coding practices and limitations of ICPC-2 diagnostic system. Standardisation of diagnostic coding, improved data linkage, and routine data validation are essential to enhance primary care data reliability for cancer surveillance, research and policy.
Carroll et al. (Fri,) studied this question.