Drug and natural health product data collection and curation in the Canadian Longitudinal Study on Aging (CLSA)

Abstract Purpose The mapping of drug and natural health product (NHP) data to standardized terminologies is central to its analysis. This study aimed to develop an efficient data collection and curation process for all drug and NHP used by Canadian Longitudinal Study on Aging (CLSA) participants. Methods The 3-step sequential data collection and curation process consisted of: 1) mapping drug inputs to the Health Canada Drug Product Database (DPD), 2) algorithm-recoding of unmapped drug and NHP inputs, and 3) manual recoding. A gold standard manually recoded input was established by two pharmacy technicians. The proportion of algorithm-correctly recoded inputs was calculated as the number of algorithm-correctly recoded inputs, based on the gold standard, divided by the number of algorithm-recoded inputs. Results Among the 30,097 CLSA Comprehensive cohort participants, 26,000 (86.4%) were using a drug or a NHP with a mean of 5.3 (SD 3.8) inputs per participant-user for a total of 137,366 inputs. Of those inputs, 70,177 (51.1%) were mapped to the Health Canada DPD, 20,729 (15.1%) were recoded by algorithms and 44,108 (32.1%) were manually recoded. In a validation sample (n = 1407 inputs), the Direct algorithm correctly classified 99.4% of drug and 99.5% of NHP inputs for which a gold standard could be established. In another validation sample of 329 manually recoded free-text inputs, consensus was reached by 2 recoders for 89.7% of drug and 74.8% of NHP inputs. Conclusion We developed an efficient 3-step process for drug and NHP data collection and curation for use in a longitudinal cohort.