Purpose: To define and validate a case-finding algorithm to identify incident colorectal cancer (CRC) in the Spanish primary care database BIFAP.
Methods: All potential incident CRC cases recorded during the study period 2001 to 2014 among patients 20 to 89 years old were identified using a defined case-finding algorithm tailored to BIFAP database characteristics and based on codes plus text mining strategies. Potential CRC cases identified by the algorithm were classified into eight homogeneous groups according to recording characteristics. Random samples of 100 cases per group were obtained, and electronic medical records were manually reviewed by two independent researchers. Positive predictive values (PPVs) were estimated per each group and for the whole sample taking into account the stratified sampling. Standardized incidence rate (SIR) of CRC was estimated and compared with that reported by the National Cancer Registry. Negative predictive value (NPV) was also estimated in a random sample of 100 non-CRC patients by the algorithm.
Results: A total of 17 008 potential CRC cases were identified. Most of them (14793; 87%) were recorded as incident diagnosis with linked clinical notes as free text, having this group a PPV of 92.1% (95%CI: 87.1%-95.3%). The overall PPV including all groups was 87.3% (95%CI: 83.3%-90.4%). SIR of CRC was 55.5 per 100.000 person-years. SIR increased with age and was higher in men as compared with women (77.7 vs 38.1 per 100.000 py, respectively) which were in line with those reported by the Network of Cancer Registries in Spain. NPV was of 100% (96.3%-100%).
Conclusions: This study shows a high validity of the CRC cases identified by the algorithm and a high level of CRC recording in BIFAP database and supports its appropriateness to validly identify incident CRC cases in BIFAP.
Keywords: BIFAP; colorectal neoplasms; electronic health care records; pharmacoepidemiology; validation.
© 2018 John Wiley & Sons, Ltd.