(183) Validation study of the death information registered in BIFAP electronic health care records data source (Spain).

Background: The validity of an electronic health records (EHR) data source to be used for epidemiological studies depends on the quality of the information available. Valid information of interest can be stored in different databanks.
Valid vital status information (dead/alive) and the corresponding dates are important to avoid time-related biases. In addition mortality and its cause are outcomes commonly used in epidemiological studies.

Objectives: To evaluate the validity of death registration in BIFAP Data source.

Methods: BIFAP is an EHR population-based data source that includes information of 10 Autonomous Regions (AR) of Spain registered in Primary care (PC) databank. Patient information registered in other databanks is also available.
Death information in BIFAP is included in the PC databank. Information of the vital status and death date is updated in the AR through the linkage with other administrative registries, but the cause of death is not available to be included.
In Spain, a death certificate is filled out by a physician including administrative information and the causes of death coded using ICD10. Computerized information is stored in the National Death Registry (NDR) database.
To evaluate the validity of death information recorded in BIFAP, patients aged 35-80 years registered as dead in BIFAP between 2005 and 2010 were included in the study. Patients not registered as dead in BIFAP, but with a PCP pattern of visits (i.e no visits in the last year of follow-up and 2 or more visits in the previous year) suggesting that they might be dead, were also included (Possible dead not registered in BIFAP).
Validation of both groups of patients was performed by linking them with the NDR database as gold standard using common personal information available in both Data Sources. Related procedures were handled by the AR and NDR Data Custodians to keep confidentiality.
The proportion of death cases in BIFAP identified as death in NDR (positive predictive value) and the concordance between the death dates were evaluated. The proportion of possible dead not registered in BIFAP identified as dead in NDR were also estimated.

Results: 97.4% of the 232,742 patients registered as death in BIFAP were also identified in the NDR. The proportion of cases with consistent information of the date of death between BIFAP and NDR was of 33.6%. This proportion increases to 84.8%, if time window increases to 0-30 days.
The 24.7% of the 12,527 possible dead not registered in BIFAP (n=3.100) were identified in the NDR.

Conclusions: The validity of death registration in BIFAP is high but the accuracy of the death date is low. The information of the cause of death in BIFAP is scarce but available in NDR.
NDR should be included as databank providing information to increase the validity and completeness of BIFAP.