Improved incidence estimates from linked vs. stand-alone electronic health records

OBJECTIVE: Electronic health records are widely used for public health research, and linked data sources are increasingly available. The added value of using linked records over stand-alone data has not been quantified for common conditions such as community-acquired pneumonia (CAP). STUDY DESIGN AND SETTING: Our cohort comprised English patients aged ≥65 years from the Clinical Practice Research Datalink, eligible for record linkage to Hospital Episode Statistics. Stand-alone general practice (GP) records were used to calculate CAP incidence over time using population-averaged Poisson regression. Incidence was then recalculated for the same patients using their linked GP-hospital admission data. Results of the two analyses were compared. RESULTS: Over 900,000 patients were included in each analysis. Population-averaged CAP incidence was 39% higher using the linked data than stand-alone data. This difference grew over time from 7% in 1997 to 83% by 2010. An increasingly larger number of pneumonia events were recorded in the hospital admission data compared to the GP data over time. CONCLUSION: Use of primary or secondary care data in isolation may not give accurate incidence estimates for important infections in older populations. Further work is needed to establish the extent of this finding in other diseases, age groups, and populations.