Association between co-morbidities and prescribed drugs in obstructive sleep apnea suspected patients: an inductive rule learning approach (Preprint)

Abstract

One way to partially impute missing clinical variables is to find associations with other informative variables described in electronic health records, as obstructive sleep apnea (OSA) has multiple clinical presentations.

To explore disease-drug associations in obstructive sleep apnea (OSA) suspected patients to improve missed diagnoses and clinical data completeness in electronic health records.

We conducted a retrospective study from a cohort of adult patients referred to the Sleep Laboratory of the University Hospital Center of São João who performed in-laboratory polysomnography. Inclusion criteria were age above eighteen years old, with a suspicion of OSA and having undergone polysomnography between January 2011 and December 2019.

A total of 481 patients taking any drug were included, resulting in 29 disease-drug strong association rules. The prescribed drugs were related to the alimentary tract and metabolism (A), cardiovascular (C), and nervous system (N). Three strong rules were obtained, describing the relationships between A10 (drugs used in diabetes) and diabetes (Lift: 2.05, Confidence: 91%), C10 (lipid modifying agents) and dyslipidemia (L: 1.28; C: 87%), and C09 (agents acting on the renin-angiotensin system) and arterial hypertension (L: 1.24; C: 95%).

We found three strong disease-drug associations rules in OSA suspected patients that can help to improve missed diagnosis in 4%, 2%, and 1% in diabetes, arterial hypertension, and dyslipidemia, respectively, based on three 2nd level ATC-codes (A10, C09, and C10).