Association rules method and big data: Evaluating frequent medication combinations associated with fractures in older adults

BACKGROUND: The association rules method is a novel methodology to ascertain patterns of medication use and combinations associated with adverse drug events.

OBJECTIVES: The aim of this case-crossover study was to apply the association rules method to ascertain medication combinations contributing to the risk of fractures in older adults.

METHODS: A nationwide representative sample of New Zealanders aged ≥65 years was sourced from the pharmaceutical collection. The first-time coded diagnosis of fracture was extracted from the National Minimum Dataset. Association rule method is a data mining technique that can be used to quickly traverse big datasets to identify a combination of items that co-occur. The association rules method were applied to identify frequent 11 medication combinations in the case and the control periods (1-14 days as hazard period, with 35-day washout period), and the association of fractures with each frequent medication combination were tested by computing a matched odd ratio (OR) and its 95% CI.

RESULTS: We identified a total of 72 184 individuals (mean age 81.5 years) from 2005 to 2014 with incident fracture and exposed to at least 1 medication of interest. The association rules method revealed codeine phosphate (aOR = 11.50, 95% CI, 7.09-15.20, concomitantly with ibuprofen), zopiclone (aOR = 2.34, 95% CI, 1.49-3.67, concomitantly with morphine) and quetiapine (OR = 1.95, 95% CI, 1.28-2.98, concomitantly with zopiclone) were associated with fractures.

CONCLUSION: The association rules method identified medication exposure combinations containing psychotropic medications and codeine are frequently associated with fractures. This novel methodology applied to big data can be an important tool to ascertain medication combinations associated with adverse drug events.