Discovering opioid users’ medical comorbidities: a data mining approach

Background: To combat the opioid crisis, scholars have investigated medical comorbidities associated with opioid use; however, the findings are often contradictory. The main problem resides in the lack of controlling for polydrug use, as the combined use of drugs can cause additive and/or synergistic effects. Methods: This study employed the apriori association rule mining algorithm, which has the capability to discover direct associations between opioid use and its comorbidities and further identify new medical comorbidities buried in the dataset as this method can process thousands of variables. Results: After controlling for polydrug use, findings show that sole opioid use associates with high systolic and diastolic blood pressures and high HbA1c, but the combined use of opioids and benzodiazepine or marijuana did not elevate systolic or diastolic blood pressure. Additionally, by including every variable in the database, this study discovered new medical comorbidities such as elevated red blood cell and gastrointestinal problems, which have not been reported in existing studies. Conclusions: The proposed analytical strategy made significant steps toward resolving the conflicting findings, as the combined use can have additive and/or synergistic effects on the medical comorbidities from opioid use. The newly discovered medical comorbidities offer future research topics. Abbreviations: BZD: benzodiazepine; MRN: marijuana; BPS: blood pressure systolic; BPD: blood pressure diastolic; AST: aspartate aminotransferase; ALT: alanine transaminase; BMI: body mass index; RBC: red blood cell count; BUN: blood urea nitrogen; MPV: mean platelet volume; EMR: electronic medical records; IRB: Institutional Review Board; CHFDW: Cerner HealthFacts® Data Warehouse.