Using record linkage to validate notification and laboratory data for a more accurate assessment of notifiable infectious diseases

BACKGROUND: Infectious disease burden is commonly assessed using notification data. Using retrospective record linkage in Western Australia, we described how well notification data captures laboratory detections of influenza, pertussis and invasive pneumococcal disease (IPD). METHODS: We linked data from the Western Australian Notifiable Infectious Diseases Database (WANIDD) and the PathWest Laboratory Database (PathWest) pertaining to the Triple I birth cohort, born in Western Australia in 1996–2012. These were combined to calculate the number of unique cases captured in each dataset alone or in both datasets. To assess the impact of under-ascertainment, we compared incidence rates calculated using WANIDD data alone and using combined data. RESULTS: Overall, there were 5550 influenza, 513 IPD (2001–2012) and 4434 pertussis cases (2000–2012). Approximately 2% of pertussis and IPD cases and 7% of influenza cases were solely recorded in PathWest. Notification of influenza and pertussis cases to WANIDD improved over time. Overall incidence rates of influenza in children aged <5 years using both datasets was 10% higher than using WANIDD data alone (IRR = 1.1, 95% CI = 1.1–1.2). CONCLUSIONS: This is the first time WANIDD data have been validated against routinely collected laboratory data. We anticipated all cases would be captured in WANIDD but found additional laboratory-confirmed cases that were not notified. Studies investigating pathogen-specific infectious disease would benefit from using multiple data sources. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (doi:10.1186/s12911-017-0484-7) contains supplementary material, which is available to authorized users.